3D edge reconstruction from posed multi-view images remains a critical yet under-explored task. While 3D Gaussian Splatting (3DGS)-based methods have recently achieved promising performance, they face two main challenges. First, edges exhibit clear discontinuities from the background, but the intrinsic smoothness of Gaussian kernels makes it challenging to model such discontinuities. Second, due to the absence of multi-view edge annotations, models are trained with pseudo labels instead. These pseudo labels extracted by pre-trained$2 D$edge detectors often exhibit cross-view inconsistencies, leading to degraded performance. To address these issues, we propose a novel uncertainty-aware 3D edge reconstruction using Difference of Gaussians (DoG) as kernels, called EdgeDoG. First, we incorporate DoG kernels to model edge discontinuities explicitly. Second, we design a dual-uncertainty strategy: primitive-level uncertainty is estimated via multi-view Fisher information to eliminate noisy 3D primitives, while pixel-level uncertainty is computed from gradients of rendered depth maps to reweight the training loss, thereby compensating for inconsistent 2D pseudo labels with robust 3D geometric cues. Extensive experiments on diverse datasets demonstrate that our method achieves superior performance compared to previous approaches.

Spatial augmented reality (SAR) directly projects digital content onto physical scenes using projectors, creating immersive experience without head-mounted displays. However, for SAR to support intelligent interaction, such as reasoning about the scene or answering user queries, it must semantically distinguish between the physical scene and the projected content. Standard Vision Language Models (VLMs) struggle with this virtual-physical ambiguity, often confusing the two contexts. To address this issue, we introduce ProCap, a novel framework that explicitly decouples projected content from physical scenes. ProCap employs a two-stage pipeline: first it visually isolates virtual and physical layers via automated segmentation; then it uses region-aware retrieval to avoid ambiguous semantic context due to projection distortion. To support this, we present RGBP (RGB + Projections), the first large-scale SAR semantic benchmark dataset, featuring 65 diverse physical scenes and over 180,000 projections with dense, decoupled annotations. Finally, we establish a dual-captioning evaluation protocol using task-specific tokens to assess physical scene and projection descriptions independently. Our experiments show that ProCap provides a robust semantic foundation for future SAR research. The source code, pre-trained models and the RGBP dataset are available on the project page: https://ZimoCao.github.io/ProCap/.

We propose UAD, an end-to-end framework with Unsupervised pretext task for vision-based Autonomous Driving, achieving the best open-loop evaluation performance in nuScenes, meanwhile showing robust closed-loop driving quality in CARLA. Our motivation stems from the observation that current end-to-end autonomous driving (E2EAD) models still mimic the modular architecture in typical driving stacks, with carefully designed supervised perception and prediction subtasks to provide environment information for oriented planning. Although achieving groundbreaking progress, such design has certain drawbacks: 1) preceding subtasks require massive high-quality 3D annotations as supervision, posing a significant impediment to scaling the training data; and 2) each submodule entails substantial computation overhead in both training and inference. To this end, we propose UAD, an E2EAD framework with an unsupervised proxy to address all these issues. Firstly, we design a novel Angular Perception Pretext to eliminate the annotation requirement. The pretext perceives the driving scene by predicting the angular-wise spatial objectness and temporal dynamics, without manual annotation. Secondly, a self-supervised training strategy, which learns the consistency of the predicted trajectories under different augment views, is proposed to enhance the planning robustness in steering scenarios. Our UAD achieves 38.7% relative improvements over UniAD on the average collision rate of nuScenes open-loop evaluation and obtains the route completion score of 98.5% in closed-loop evaluation of CARLA's Town05 Long benchmark, which outperforms the recent work VADv2. Moreover, the proposed method consumes only 44.3% training resources of UniAD and runs 3.4× faster in inference when employing the same backbone network. Our innovative design not only for the first time demonstrates unarguable performance advantages over supervised counterparts, but also enjoys unprecedented efficiency in data, training, and inference.

Detecting occlusion edges from videos is a critical yet under-explored task, with recent works focusing on single-image occlusion edge detection while ignoring dynamic patterns in videos. In videos, occlusion edges typically occur when a moving object occludes either the background or another object, resulting in two types of occlusion edges:object-background (OB) edgesthat are characterized by appearance contrast and motion, and object-object (OO) edgesthat face ambiguity. Inspired by these observations, we propose a novel Motion and Spatio-Temporal Aggregated Network (MaSTAN) and treat the two edge types differently for more effectively detecting occlusion edges in videos. Specifically, we first extract spatial semantics and motion patterns from the videos and propose a novel Temporal Feature Propagation module (TFP) for temporal cue aggregation. Next, we put forward a Dual-branch Gated-attention Decoder (DG-Decoder) to generate edge-specific features for predicting the final occlusion edge maps. Extensive experiments on the OVIS-OE benchmark, the first large benchmark dedicated to video occlusion edge detection, demonstrate that MaSTAN achieves state-of-the-art performance, significantly advancing the capability of occlusion edge detection in video. The source code and benchmark will be made publicly available.

Masked autoencoders have been extended beyond images, but random masking often fails to capture non-uniform motion regions in inherently disordered and irregular data like point cloud videos. In this paper, we propose a motion-guided disentanglement method (MGD) to improve masked autoencoders for point cloud video representation learning. Specifically, we begin by estimating motion intensity using an optimal transport approach, which guides the separate masking of dynamic and static regions. This motion-guided masking ensures balanced coverage, addressing the limitations of random masking in capturing non-uniformly distributed motion regions. Furthermore, we disentangle the prediction tasks into motion prediction for high-motion point tubes and appearance reconstruction for low-motion ones. This disentanglement enables the model to more effectively capture both motion and appearance in point cloud videos. We conducted experiments on four widely used point cloud video datasets—NTU RGB+D, MSR-Action3D, NvGesture, and SHREC’17—which demonstrate that our approach consistently improves masked autoencoders for point cloud video representation learning, achieving new state-of-the-art results. Code will be publicly available on GitHub.

Projector compensation seeks to correct geometric and photometric distortions that occur when images are projected onto nonplanar or textured surfaces. However, most existing methods are highly setup-dependent, requiring fine-tuning or retraining whenever the surface, lighting, or projector-camera pose changes. Progress has been limited by two key challenges: (1) the absence of large, diverse training datasets and (2) existing geometric correction models are typically constrained by specific spatial setups; without further retraining or fine-tuning, they often fail to generalize directly to novel geometric configurations. We introduce SIComp, the first Setup-Independent framework for full projector Compensation, capable of generalizing to unseen setups without fine-tuning or retraining. To enable this, we construct a large-scale real-world dataset spanning 277 distinct projector-camera setups. SIComp adopts a co-adaptive design that decouples geometry and photometry: A carefully tailored optical flow module performs online geometric correction, while a novel photometric network handles photometric compensation. To further enhance robustness under varying illumination, we integrate intensity-varying surface priors into the network design. Extensive experiments demonstrate that SIComp consistently produces high-quality compensation across diverse unseen setups, substantially outperforming existing methods in terms of generalization ability and establishing the first generalizable solution to projector compensation. The code and dataset are available on our project page: https://hai-bo-li.github.io/SIComp/.

Projector photometric compensation corrects color distortions introduced by surface texture, reflection, and ambient lighting. Existing deep learning-based methods usually require professional scene-specific data collection and lack consideration for perceptual quality. To address this limitation, we present a diffusion-based photometric compensation method that reconstructs compensation images under photometric and content-aware guidance. Specifically, we fi rst mo del th e ph otometric distortions introduced during projection as environment-dependent additive noise, thereby reformulating the photometric compensation problem as a denoising task with physical constraints. Next, we introduce a diffusion model, which generates compensation images by following an additive trajectory to iteratively remove the noise. Finally, to accurately estimate the noise at each timestep, by analyzing the factors that contribute to distortions in the physical process of projection and capturing, we design a noise estimation network that incorporates features of both photometry-aware and content conditions. Experiments show that our method achieves superior visual performance in unknown scenarios, thereby exhibiting significant practical advantages over prior art. Our source code is available at https://github.com/cyxwang/DiffPC.

Text labels are widely used to convey auxiliary information in visualization and graphic design. The substantial variability in the categories and structures of labeled objects leads to diverse label layouts. Recent single-model learning-based solutions in label placement struggle to capture fine-grained differences between these layouts, which in turn limits their performance. In addition, although human designers often consult previous works to gain design insights, existing label layouts typically serve merely as training data, limiting the extent to which embedded design knowledge can be exploited. To address these challenges, we propose a mixture of cluster-guided experts (MoCE) solution for label placement. In this design, multiple experts jointly refine layout features, with each expert responsible for a specific cluster of layouts. A cluster-based gating function assigns input samples to experts based on representation clustering. We implement this idea through the Label Placement Cluster-guided Experts (LPCE) model, in which a MoCE layer integrates multiple feed-forward networks (FFNs), with each expert composed of a pair of FFNs. Furthermore, we introduce a retrieval augmentation strategy into LPCE, which retrieves and encodes reference layouts for each input sample to enrich its representations. Extensive experiments demonstrate that LPCE achieves superior performance in label placement, both quantitatively and qualitatively, surpassing a range of state-of-the-art baselines. Our algorithm is available at https://github.com/PingshunZhang/LPCE.

Estimating blood flow speed is essential in many medical and physiological applications, yet it is extremely challenging due to complex vascular structure and flow dynamics, particularly for cerebral cortex regions. Existing techniques, such as Optical Doppler Tomography (ODT), generally require complex hardware control and signal processing, and still suffer from inherent system-level artifacts. To address these challenges, we propose a new learning-based approach named OCTA-Flow, which directly estimates vascular blood flow speed from Optical Coherence Tomography Angiography (OCTA) images that are commonly used for vascular structure analysis. OCTA-Flow employs several novel components to achieve this goal. First, using an encoder-decoder architecture, OCTA-Flow leverages ODT data as pseudo labels during training, thus bypassing the difficulty of collecting ground truth data. Second, to capture the relationship between vessels of varying scales and their flow speed, we design an Adaptive Window Fusion module that employs multiscale window attention. Third, to mitigate ODT artifacts, we incorporate a Conditional Random Field Decoder that promotes smoothness and consistency in the estimated blood flow. Together, these innovations enable OCTA-Flow to effectively produce accurate flow estimation, suppress the artifacts in ODT, and enhance practicality, benefiting from the established techniques of OCTA data acquisition. The code and data are available at https://github.com/Spritea/OCTA-Flow.

The emergence of Vision-Language Models (VLMs) represents a significant advancement in integrating computer vision with Large Language Models (LLMs) to generate detailed text descriptions from visual inputs. Despite their growing importance, the security of VLMs, particularly against backdoor attacks, is under explored. Moreover, prior works often assume attackers have access to the original training data, which is often unrealistic. In this paper, we address a more practical and challenging scenario where attackers must rely solely on Out-Of-Distribution (OOD) data. We introduce VLOOD (Backdoor Vision-Language Models using Out-of-Distribution Data), a novel approach with two key contributions: (1) demonstrating backdoor attacks on VLMs in complex image-to-text tasks while minimizing degradation of the original semantics under poisoned inputs, and (2) proposing innovative techniques for backdoor injection without requiring any access to the original training data. Our evaluation on image captioning and visual question answering (VQA) tasks confirms the effectiveness of VLOOD, revealing a critical security vulnerability in VLMs and laying the foundation for future research on securing multimodal models against sophisticated threats.

Deep learning has achieved significant success by training on balanced datasets. However, real-world data often exhibit long-tailed distributions. Empirical studies have revealed that long-tailed data skew data representations, where head classes dominate the feature space. Many methods have been proposed to empirically rectify the skewed representations. However, a clear understanding of the underlying cause and extent of this skew remains lacking. In this study, we provide a comprehensive theoretical analysis to elucidate how long-tailed data affect feature distributions, deriving the conditions under which centers of tail classes shrink together or even collapse into a single point. This results in overlapping feature distributions of tail classes, making features in the overlapping regions inseparable. Moreover, we demonstrate that merely empirically correcting the skewed representations of the training data is insufficient to separate the overlapping features due to distribution shifts between the training and real data. To address these challenges, we propose a novel long-tailed representation learning method, FeatRecon. It reconstructs the feature space in order to arrange features from different classes into symmetricial and linearly separable regions. This, in turn, enhances the model’s robustness to long-tailed data. We validate the effectiveness of our method through extensive experiments on the CIFAR-10-LT, CIFAR-100-LT, ImageNet-LT, and iNaturalist 2018 datasets.

Projector-Based adversarial attack aims to project carefully designed light patterns (i.e., adversarial projections) onto scenes to deceive deep image classifiers. It has potential applications in privacy protection and the development of more robust classifiers. However, existing approaches primarily focus on individual classifiers and fixed camera poses, often neglecting the complexities of multi-classifier systems and scenarios with varying camera poses. This limitation reduces their effectiveness when introducing new classifiers or camera poses. In this paper, we introduce Classifier-Agnostic Projector-Based Adversarial Attack (CAPAA) to address these issues. First, we develop a novel classifier-agnostic adversarial loss and optimization framework that aggregates adversarial and stealthiness loss gradients from multiple classifiers. Then, we propose an attention-based gradient weighting mechanism that concentrates perturbations on regions of high classification activation, thereby improving the robustness of adversarial projections when applied to scenes with varying camera poses. Our extensive experimental evaluations demonstrate that CAPAA achieves both a higher attack success rate and greater stealthiness compared to existing baselines. Codes are available at: https://github.com/ZhanLiQxQ/CAPAA.

3D Geometric Graph Neural Networks (GNNs) have emerged as transformative tools for modeling molecular data. Despite their predictive power, these models often suffer from limited interpretability, raising concerns for scientific applications that require reliable and transparent insights. While existing methods have primarily focused on explaining molecular substructures in 2D GNNs, the transition to 3D GNNs introduces unique challenges, such as handling the implicit dense edge structures created by a cutoff radius. To tackle this, we introduce a novel explanation method specifically designed for 3D GNNs, which localizes the explanation to the immediate neighborhood of each node within the 3D space. Each node is assigned an radius of influence, defining the localized region within which message passing captures spatial and structural interactions crucial for the model's predictions. This method leverages the spatial and geometric characteristics inherent in 3D graphs. By constraining the subgraph to a localized radius of influence, the approach not only enhances interpretability but also aligns with the physical and structural dependencies typical of 3D graph applications, such as molecular learning.

Federated stochastic bilevel optimization has been actively studied in recent years due to its widespread applications in machine learning. However, most existing federated stochastic bilevel optimization algorithms require the computation of second-order Hessian and Jacobian matrices, which leads to longer running times in practice. To address these challenges, we propose a novel federated stochastic variance-reduced bilevel gradient descent algorithm that relies solely on first-order oracles. Specifically, our approach does not require the computation of second-order Hessian and Jacobian matrices, significantly reducing running time. Furthermore, we introduce a novel learning rate mechanism, i.e., a constant single-time-scale learning rate, to coordinate the update of different variables. We also present a new strategy to establish the convergence rate of our algorithm. Finally, the extensive experimental results confirm the efficacy of our proposed algorithm.

Localizing predefined 3D keypoints in a 2D image is an effective way to establish 3D-2D correspondences for instance-level 6DoF object pose estimation. However, unreliable localization results of invisible keypoints degrade the quality of correspondences. In this paper, we address this issue by localizing the important keypoints in terms of visibility. Since keypoint visibility information is currently missing in the dataset collection process, we propose an efficient way to generate binary visibility labels from available object-level annotations, for keypoints of both asymmetric objects and symmetric objects. We further derive real-valued visibility-aware importance from binary labels based on the PageRank algorithm. Taking advantage of the flexibility of our visibility-aware importance, we construct VAPO (Visibility-Aware POse estimator) by integrating the visibility-aware importance with a state-of-the-art pose estimation algorithm, along with additional positional encoding. VAPO can work in both CAD-based and CAD-free settings. Extensive experiments are conducted on popular pose estimation benchmarks including Linemod, Linemod-Occlusion, and YCB-V, demonstrating that VAPO clearly achieves state-of-the-art performances. Project page: https://github.com/RuyiLian/VAPO.

Underwater image restoration aims to remove geometric and color distortions due to water refraction, absorption, and scattering. Previous studies focus on restoring either color or geometry, but to our best knowledge, not both. However, in practice it may be cumbersome to address the two rectifications one by one. In this paper, we propose NeuroPump, a self-supervised method to simultaneously optimize and rectify underwater geometry and color as if water were pumped out. The key idea is to explicitly model refraction, absorption, and scattering in Neural Radiance Field (NeRF) pipeline, such that it not only performs simultaneous geometric and color rectification, but also enables to synthesize novel views and optical effects by controlling the decoupled parameters. In addition, to address the lack of real paired ground truth images, we propose an underwater 360 benchmark dataset that has real paired (i.e., with and without water) images. Our method clearly outperforms other baselines both quantitatively and qualitatively. Our code and dataset is available at https://ygswu.github.io/NeuroPump.github.io/.

Transformer-based algorithms, such as LoRAT, have significantly enhanced object-tracking performance. However, these approaches rely on a standard attention mechanism, which incurs quadratic token complexity, making real-time inference computationally expensive. In this paper, we introduce LoRATv2, a novel tracking framework that addresses these limitations with three main contributions. First, LoRATv2 integrates frame-wise causal attention, which ensures full self-attention within each frame while enabling causal dependencies across frames, significantly reducing computational overhead. Moreover, key-value (KV) caching is employed to efficiently reuse past embeddings for further speedup. Second, building on LoRAT's parameter-efficient fine-tuning, we propose Stream-Specific LoRA Adapters (SSLA). As frame-wise causal attention introduces asymmetry in how streams access temporal information, SSLA assigns dedicated LoRA modules to the template and each search stream, with the main ViT backbone remaining frozen. This allows specialized adaptation for each stream's role in temporal tracking. Third, we introduce a two-phase progressive training strategy, which first trains a single-search-frame tracker and then gradually extends it to multi-search-frame inputs by introducing additional LoRA modules. This curriculum-based learning paradigm improves long-term tracking while maintaining training efficiency. In extensive experiments on multiple benchmarks, LoRATv2 achieves state-of-the-art performance, substantially improved efficiency, and a superior performance-to-FLOPs ratio over state-of-the-art trackers. The code is available at https://github.com/LitingLin/LoRATv2.

Self-supervised learning plays an important role in current state-of-the-art semi-supervised learning (SSL) methods. These methods learn inter-class heterogeneity among data and generate pseudo-labels based on class level representations. However, they often neglect intra-class heterogeneity, resulting in the under-exploitation of finer-grained semantic relationships within classes. To address this limitation, we introduce PivotAlign, a novel SSL approach that aims to 1) learn hierarchical representations to detect both interclass and intra-class semantic relationships, and 2) refine pseudo-labels based on learned representations with a class-debiasing strategy. Specifically, we first learn a set of pivots as sub-prototypes of classes. We then train representations so that features align with the assigned pivot and are hierarchically grouped based on both inter-class and intra-class heterogeneity. This allows us to capture both inter-class and intra-class semantic relationships among data and leverage them to better assign and refine pseudo-labels. Additionally, since SSL methods are prone to bias toward classes that are easier to learn, we further re-balance class predictions to alleviate this class bias. We demonstrate the effectiveness of PivotAlign on various SSL benchmarks, where PivotAlign achieves state-of-the-art performances. The source code will be released upon publication of the work.

Monocular 3D object detection plays a crucial role In the field of self-driving cars, estimating the size and location of objects solely based on input images. However, a notable disparity exists between the training and inference of 3D object detectors. This discrepancy arises because during inference, monocular 3D detectors depend solely on images captured by cameras; while during training, these methods require 3D ground truths labeled on point cloud data, which is obtained using specialized devices like LiDAR. This discrepancy creates a break in the data loop, preventing the feedback data from production cars from being utilized to enhance the robustness of the detectors. To address this issue and establish a connection in the data loop, we present a weakly-supervised solution that trains monocular 3D object detectors solely using 2D labels, eliminating the requirement for 3D ground truths. Our approach considers two view consistency: spatial and temporal view consistency, which play a crucial role in regulating the prediction of 3D bounding boxes. Spatial view consistency is achieved by employing projection and multi-view consistency techniques to guide the optimization of the target's location and size. We leverage temporal viewpoint consistency to provide temporal multi-view image pairs, and we further introduce temporal movement consistency to tackle the challenge of dynamic scenes. With only 2D ground truths, our method achieves comparable performance to fully supervised methods. Additionally, our method can be employed as a pre-training method and achieves significant improvement when fine-tuned with a small proportion of fully supervised labels.

Most previous learning-based graph matching algorithms solve the quadratic assignment problem (QAP) by dropping one or more of the matching constraints and adopting a relaxed assignment solver to obtain sub-optimal correspondences. Such relaxation may actually weaken the original graph matching problem, and in turn hurt the matching performance. In this paper, we propose a deep learning-based graph matching framework that works for the original QAP without compromising on the matching constraints. In particular, we design an affinityassignment prediction network to jointly learn the pairwise affinity and estimate the node assignments, and we then develop a differentiable solver inspired by the probabilistic perspective of the pairwise affinities. Aiming to obtain better matching results, the probabilistic solver refines the estimated assignments in an iterative manner to impose both discrete and one-to-one matching constraints. The proposed method is trained in a supervised manner, evaluated on several benchmarks related to semantic keypoint corresponding, matching of social networks and pure QAP instances. In all experiment, it exhibits state-of-the-art matching performance on all benchmarks.

We propose LAPIG, a language guided projector image generation method with surface adaptation and stylization. LAPIG consists of a projector-camera system and a target textured projection surface. LAPIG takes the user text prompt as input and aims to transform the surface style using the projector. LAPIG's key challenge is that due to the projector's physical brightness limitation and the surface texture, the viewer's perceived projection may suffer from color saturation and artifacts in both dark and bright regions, such that even with the state-of-the-art projector compensation techniques, the viewer may see clear surface texture-related artifacts. Therefore, how to generate a projector image that follows the user's instruction while also displaying minimum surface artifacts is an open problem. To address this issue, we propose projection surface adaptation (PSA) that can generate compensable surface stylization. We first train two networks to simulate the projector compensation and project-and-capture processes, this allows us to find a satisfactory projector image without real project-and-capture and utilize gradient descent for fast convergence. Then, we design content and saturation losses to guide the projector image generation, such that the generated image shows no clearly perceivable artifacts when projected. Finally, the generated image is projected for visually pleasing surface style morphing effects. The source code and more results are available on the project page: https://Yu-chen-Deng.github.io/LAPIG/.

We present GS-ProCams, the first Gaussian Splatting-based framework for projector-camera systems (ProCams). GS-ProCams is not only view-agnostic but also significantly enhances the efficiency of projection mapping (PM) that requires establishing geometric and radiometric mappings between the projector and the camera. Previous CNN-based ProCams are constrained to a specific viewpoint, limiting their applicability to novel perspectives. In contrast, NeRF-based ProCams support view-agnostic projection mapping, however, they require an additional co-located light source and demand significant computational and memory resources. To address this issue, we propose GS-ProCams that employs 2D Gaussian for scene representations, and enables efficient view-agnostic ProCams applications. In particular, we explicitly model the complex geometric and photometric mappings of ProCams using projector responses, the projection surface's geometry and materials represented by Gaussians, and the global illumination component. Then, we employ differentiable physically-based rendering to jointly estimate them from captured multi-view projections. Compared to state-of-the-art NeRF-based methods, our GS-ProCams eliminates the need for additional devices, achieving superior ProCams simulation quality. It also uses only 1/10 of the GPU memory for training and is 900 times faster in inference speed. Please refer to our project page for the code and dataset: https://realqingyue.github.io/GS-ProCams/.

Projector-camera systems (ProCams) simulation aims to model the physical project-and-capture process and associated scene parameters of a ProCams, and is crucial for spatial augmented reality (SAR) applications such as ProCams relighting and projector compensation. Recent advances use an end-to-end neural network to learn the project-and-capture process. However, these neural network-based methods often implicitly encapsulate scene parameters, such as surface material, gamma, and white balance in the network parameters, and are less interpretable and hard for novel scene simulation. Moreover, neural networks usually learn the indirect illumination implicitly in an image-to-image translation way which leads to poor performance in simulating complex projection effects such as soft-shadow and interreflection. In this paper, we introduce a novel path tracing-based differentiable projector-camera systems (DPCS), offering a differentiable ProCams simulation method that explicitly integrates multi-bounce path tracing. Our DPCS models the physical project-and-capture process using differentiable physically-based rendering (PBR), enabling the scene parameters to be explicitly decoupled and learned using much fewer samples. Moreover, our physically-based method not only enables high-quality downstream ProCams tasks, such as ProCams relighting and projector compensation, but also allows novel scene simulation using the learned scene parameters. In experiments, DPCS demonstrates clear advantages over previous approaches in ProCams simulation, offering better interpretability, more efficient handling of complex interreflection and shadow, and requiring fewer training samples. The code and dataset are available on the project page: https://jijiangli.github.io/DPCS/.

Placing text labels is a common way to explain key elements in a given scene. Given a graphic input and original label information, how to place labels to meet both geometric and aesthetic requirements is an open challenging problem. Geometry-wise, traditional rule-driven solutions struggle to capture the complex interactions between labels, let alone consider graphical/appearance content. In terms of aesthetics, training/evaluation data ideally require nontrivial effort and expertise in design, thus resulting in a lack of decent datasets for learning-based methods. To address the above challenges, we formulate the task with a graph representation, where nodes correspond to labels and edges to interactions between labels, and treat label placement as a node position prediction problem. With this novel representation, we design a Label Placement Graph Transformer (LPGT) to predict label positions. Specifically, edge-level attention, conditioned on node representations, is introduced to reveal potential relationships between labels. To integrate graphic/image information, we design a feature aligning strategy that extracts deep features for nodes and edges efficiently. Next, to address the dataset issue, we collect commercial illustrations with professionally designed label layouts from household appliance manuals, and annotate them with useful information to create a novel dataset named the Appliance Manual Illustration Labels (AMIL) dataset. In the thorough evaluation on AMIL, our LPGT solution achieves promising label placement performance compared with popular baselines. Our algorithm and dataset are available at https://github.com/JingweiQu/LPGT.

Edge segmentation is well-known to be subjective due to personalized annotation styles and preferred granular-ity. However, most existing deterministic edge detection methods produce only a single edge map for one input image. We argue that generating multiple edge maps is more reasonable than generating a single one considering the subjectivity and ambiguity of the edges. Thus motivated, in this paper we propose multiple granularity edge detection, called MuGE, which can produce a wide range of edge maps, from approximate object contours to fine texture edges. Specifically, we first propose to design an edge granularity network to estimate the edge granularity from an individual edge annotation. Subsequently, to guide the generation of diversified edge maps, we integrate such edge granularity into the multi-scale feature maps in the spatial domain. Meanwhile, we decompose the feature maps into low-frequency and high-frequency parts, where the encoded edge granularity is further fused into the high-frequency part to achieve more precise control over the details of the produced edge maps. Compared to previous methods, MuGE is able to not only generate multiple edge maps at different controllable granularities but also achieve a com-petitive performance on the BSDS500 and Multicue benchmark datasets.

Contrastive learning is a powerful learning paradigm that can fully exploit cheap, self-supervised learning signals in the data. Crucial to its success is the choice of the positive and negative pairs, yet they are typically based on simple heuristics, and may generate low-quality or even harmful supervision. How to identify beneficial and harmful contrastive pairs? Most of the existing works make the decision based on the input samples directly, which is a static view and ignores how the learning process would respond to such learning constraints with varying level of quality and feasibility (or difficulty). In this paper, we propose instead to use the solution path of the contrastive learning process itself as a more dynamic guidance. Specifically, we use a step-wise reweighting on the contrastive loss of each sample pair, based on their representations that are being optimized in each step of the back-propagation as an instant feedback. We show that, when the re-weighting function is inversely proportional to the updated sample similarities, we can achieve a dynamic denoising of the contrastive loss, so that those "infeasible" and "accomplished" sample pairs are naturally discounted, while those "achievable yet unfinished" ones naturally emphasized. Our reweighting scheme, which we call DynaDeno (Dynamic Denoising), is a unified framework for continuously improving the allocation of the learning resources in contrastive learning, which takes into account both the different levels of uncertainty with positive and negative sample pairs, as well as how the learning process responds to such constraints. When applied to GNN-based node classification, our approach has demonstrated promising results on widely used benchmark datasets.

Although the performance of Temporal Action Segmentation (TAS) has been improved in recent years, achieving promising results often comes with a high computational cost due to dense inputs, complex model structures, and resource-intensive post-processing requirements. To improve the efficiency while keeping the high performance, we present a novel perspective centered on per-segment classification. By harnessing the capabilities of Transformers, we tokenize each video segment as an instance token, endowed with intrinsic instance segmentation. To realize efficient action segmentation, we introduce BaFormer, a boundary-aware Transformer network. It employs instance queries for instance segmentation and a global query for class-agnostic boundary prediction, yielding continuous segment proposals. During inference, BaFormer employs a simple yet effective voting strategy to classify boundary-wise segments based on instance segmentation. Remarkably, as a single-stage approach, BaFormer significantly reduces the computational costs, utilizing only 6% of the running time compared to the state-of-the-art method DiffAct, while producing better or comparable accuracy over several popular benchmarks. The code for this project is publicly available at https://github.com/peiyao-w/BaFormer.

Bird's-eye-view (BEV) map layout estimation requires an accurate and full understanding of the semantics for the environmental elements around the ego car to make the results coherent and realistic. Due to the challenges posed by occlusion, unfavourable imaging conditions and low resolution, \emph{generating} the BEV semantic maps corresponding to corrupted or invalid areas in the perspective view (PV) is appealing very recently. \emph{The question is how to align the PV features with the generative models to facilitate the map estimation}. In this paper, we propose to utilize a generative model similar to the Vector Quantized-Variational AutoEncoder (VQ-VAE) to acquire prior knowledge for the high-level BEV semantics in the tokenized discrete space. Thanks to the obtained BEV tokens accompanied with a codebook embedding encapsulating the semantics for different BEV elements in the groundtruth maps, we are able to directly align the sparse backbone image features with the obtained BEV tokens from the discrete representation learning based on a specialized token decoder module, and finally generate high-quality BEV maps with the BEV codebook embedding serving as a bridge between PV and BEV. We evaluate the BEV map layout estimation performance of our model, termed VQ-Map, on both the nuScenes and Argoverse benchmarks, achieving 62.2/47.6 mean IoU for surround-view/monocular evaluation on nuScenes, as well as 73.4 IoU for monocular evaluation on Argoverse, which all set a new record for this map layout estimation task. The code and models are available on \url{https://github.com/Z1zyw/VQ-Map}.

Graph neural networks (GNNs) are recognized for their strong performance across various applications, with the backpropagation (BP) algorithm playing a central role in the development of most GNN models. However, despite its effectiveness, BP has limitations that challenge its biological plausibility and affect the efficiency, scalability and parallelism of training neural networks for graph-based tasks. While several non-backpropagation (non-BP) training algorithms, such as the direct feedback alignment (DFA), have been successfully applied to fully-connected and convolutional network components for handling Euclidean data, directly adapting these non-BP frameworks to manage non-Euclidean graph data in GNN models presents significant challenges. These challenges primarily arise from the violation of the independent and identically distributed (i.i.d.) assumption in graph data and the difficulty in accessing prediction errors for all samples (nodes) within the graph. To overcome these obstacles, in this paper we propose DFA-GNN, a novel forward learning framework tailored for GNNs with a case study of semi-supervised learning. The proposed method breaks the limitations of BP by using a dedicated forward training mechanism. Specifically, DFA-GNN extends the principles of DFA to adapt to graph data and unique architecture of GNNs, which incorporates the information of graph topology into the feedback links to accommodate the non-Euclidean characteristics of graph data. Additionally, for semi-supervised graph learning tasks, we developed a pseudo error generator that spreads residual errors from training data to create a pseudo error for each unlabeled node. These pseudo errors are then utilized to train GNNs using DFA. Extensive experiments on 10 public benchmarks reveal that our learning framework outperforms not only previous non-BP methods but also the standard BP methods, and it exhibits excellent robustness against various types of noise and attacks.

Multimodal vision-language (VL) learning has noticeably pushed the tendency toward generic intelligence owing to emerging large foundation models. However, tracking, as a fundamental vision problem, surprisingly enjoys less bonus from recent flourishing VL learning. We argue that the reasons are two-fold: the lack of large-scale vision-language annotated videos and ineffective vision-language interaction learning of current works. These nuisances motivate us to design more effective vision-language representation for tracking, meanwhile constructing a large database with language annotation for model learning. Particularly, in this paper, we first propose a general attribute annotation strategy to decorate videos in six popular tracking benchmarks, which contributes a large-scale vision-language tracking database with more than 23,000 videos. We then introduce a novel framework to improve tracking by learning a unified-adaptive VL representation, where the cores are the proposed asymmetric architecture search and modality mixer (ModaMixer). To further improve VL representation, we introduce a contrastive loss to align different modalities. To thoroughly evidence the effectiveness of our method, we integrate the proposed framework on three tracking methods with different designs, i.e., the CNN-based SiamCAR (Guo et al. 2020), the Transformer-based OSTrack (Ye et al. 2022), and the hybrid structure TransT (Chen et al. 2021). The experiments demonstrate that our framework can significantly improve all baselines on six benchmarks. Besides empirical results, we theoretically analyze our approach to show its rationality. By revealing the potential of VL representation, we expect the community to divert more attention to VL tracking and hope to open more possibilities for future tracking with diversified multimodal messages.

Deep learning-based solutions have achieved impressive performance in semantic segmentation but often require large amounts of training data with fine-grained annotations. To alleviate such requisition, a variety of weakly supervised annotation strategies have been proposed, among which scribble supervision is emerging as a popular one due to its user-friendly annotation way. However, the sparsity and diversity of scribble annotations make it nontrivial to train a network to produce deterministic and consistent predictions directly. To address these issues, in this paper we propose holistic solutions involving the design of network structure, loss and training procedure, named CC4S to improve Certainty and Consistency for Scribble-Supervised Semantic Segmentation. Specifically, to reduce uncertainty, CC4S embeds a random walk module into the network structure to make neural representations uniformly distributed within similar semantic regions, which works together with a soft entropy loss function to force the network to produce deterministic predictions. To encourage consistency, CC4S adopts self-supervision training and imposes the consistency loss on the eigenspace of the probability transition matrix in the random walk module (we named neural eigenspace). Such self-supervision inherits the category-level discriminability from the neural eigenspace and meanwhile helps the network focus on producing consistent predictions for the salient parts and neglect semantically heterogeneous backgrounds. Finally, to further improve the performance, CC4S uses the network predictions as pseudo-labels and retrains the network with an extra color constraint regularizer. From comprehensive experiments, CC4S achieves comparable performance to those from fully supervised methods and shows promising robustness under extreme supervision cases.

We propose a conceptually novel, flexible, and effective framework (named T-Net++) for the task of two-view correspondence pruning. T-Net++ comprises two unique structures: the "-'' structure and the "|'' structure. The "-'' structure utilizes an iterative learning strategy to process correspondences, while the "|'' structure integrates all feature information of the "-'' structure and produces inlier weights. Moreover, within the "|'' structure, we design a new Local-Global Attention Fusion module to fully exploit valuable information obtained from concatenating features through channel-wise and spatial-wise relationships. Furthermore, we develop a Channel-Spatial Squeeze-and-Excitation module, a modified network backbone that enhances the representation ability of important channels and correspondences through the squeeze-and-excitation operation. T-Net++ not only preserves the permutation-equivariance manner for correspondence pruning, but also gathers rich contextual information, thereby enhancing the effectiveness of the network. Experimental results demonstrate that T-Net++ outperforms other state-of-the-art correspondence pruning methods on various benchmarks and excels in two extended tasks.

Many complex social, biological, or physical systems are characterized as networks, and recovering the missing links of a network could shed important lights on its structure and dynamics. A good topological representation is crucial to accurate link modeling and prediction, yet how to account for the kaleidoscopic changes in link formation patterns remains a challenge, especially for analysis in cross-domain studies. We propose a new link representation scheme by projecting the local environment of a link into a "dipole plane", where neighboring nodes of the link are positioned via their relative proximity to the two anchors of the link, like a dipole. By doing this, complex and discrete topology arising from link formation is turned to differentiable point-cloud distribution, opening up new possibilities for topological feature-engineering with desired expressiveness, interpretability and generalization. Our approach has comparable or even superior results against state-of-the-art GNNs, meanwhile with a model up to hundreds of times smaller and running much faster. Furthermore, it provides a universal platform to systematically profile, study, and compare link-patterns from miscellaneous real-world networks. This allows building a global link-pattern atlas, based on which we have uncovered interesting common patterns of link formation, i.e., the bridge-style, the radiation-style, and the community-style across a wide collection of networks with highly different nature.

Domain adaptive detection aims to improve the generalization of detectors on target domain. To reduce discrepancy in feature distributions between two domains, recent approaches achieve domain adaption through feature alignment in different granularities via adversarial learning. However, they neglect the relationship between multiple granularities and different features in alignment, degrading detection. Addressing this, we introduce a unified multi-granularity alignment (MGA)-based detection framework for domain-invariant feature learning. The key is to encode the dependencies across different granularities including pixel-, instance-, and category-levels simultaneously to align two domains. Specifically, based on pixel-level features, we first develop an omni-scale gated fusion (OSGF) module to aggregate discriminative representations of instances with scale-aware convolutions, leading to robust multi-scale detection. Besides, we introduce multi-granularity discriminators to identify where, either source or target domains, different granularities of samples come from. Note that, MGA not only leverages instance discriminability in different categories but also exploits category consistency between two domains for detection. Furthermore, we present an adaptive exponential moving average (AEMA) strategy that explores model assessments for model update to improve pseudo labels and alleviate local misalignment problem, boosting detection robustness. Extensive experiments on multiple domain adaption scenarios validate the superiority of MGA over other approaches on FCOS and Faster R-CNN detectors.

Egocentric action recognition is gaining significant attention in the field of human action recognition. In this paper, we address data scarcity issue in egocentric action recognition from a compositional generalization perspective. To tackle this problem, we propose a free-form composition network (FFCN) that can simultaneously learn disentangled verb, preposition, and noun representations, and then use them to compose new samples in the feature space for rare classes of action videos. First, we use a graph to capture the spatial-temporal relations among different hand/object instances in each action video. We thus decompose each action into a set of verb and preposition spatial-temporal representations using the edge features in the graph. The temporal decomposition extracts verb and preposition representations from different video frames, while the spatial decomposition adaptively learns verb and preposition representations from action-related instances in each frame. With these spatial-temporal representations of verbs and prepositions, we can compose new samples for those rare classes in a free-form manner, which is not restricted to a rigid form of a verb and a noun. The proposed FFCN can directly generate new training data samples for rare classes, hence significantly improve action recognition performance. We evaluated our method on three popular egocentric action recognition datasets, Something-Something V2, H2O, and EPIC-KITCHENS-100, and the experimental results demonstrate the effectiveness of the proposed method for handling data scarcity problems, including long-tailed and few-shot egocentric action recognition.

Recently studies have shown the potential of weakly supervised multi-object tracking and segmentation, but the drawbacks of coarse pseudo mask label and limited utilization of temporal information remain to be unresolved. To address these issues, we present a framework that directly uses box label to supervise the segmentation network without resorting to pseudo mask label. In addition, we propose to fully exploit the temporal information from two perspectives. Firstly, we integrate optical flow-based pairwise consistency to ensure mask consistency across frames, thereby improving mask quality for segmentation. Secondly, we propose a temporally adjacent pair-based sampling strategy to adapt instance embedding learning for data association in tracking. We combine these techniques into an end-to-end deep model, named BoxMOTS, which requires only box annotation without mask supervision. Extensive experiments demonstrate that our model surpasses current state-of-the-art by a large margin, and produces promising results on KITTI MOTS and BDD100K MOTS. The source code is available at https://github.com/Spritea/BoxMOTS.

Domain Generalization (DG) aims to learn a generalizable model on the unseen target domain by only training on the multiple observed source domains. Although a variety of DG methods have focused on extracting domain-invariant features, the domain-specific class-relevant features have attracted attention and been argued to benefit generalization to the unseen target domain. To take into account the class-relevant domain-specific information, in this paper we propose an Information theory iNspired diSentanglement and pURification modEl (INSURE) to explicitly disentangle the latent features to obtain sufficient and compact (necessary) class-relevant feature for generalization to the unseen domain. Specifically, we first propose an information theory inspired loss function to ensure the disentangled class-relevant features contain sufficient class label information and the other disentangled auxiliary feature has sufficient domain information. We further propose a paired purification loss function to let the auxiliary feature discard all the class-relevant information and thus the class-relevant feature will contain sufficient and compact (necessary) class-relevant information. Moreover, instead of using multiple encoders, we propose to use a learnable binary mask as our disentangler to make the disentanglement more efficient and make the disentangled features complementary to each other. We conduct extensive experiments on five widely used DG benchmark datasets including PACS, VLCS, OfficeHome, TerraIncognita, and DomainNet. The proposed INSURE achieves state-of-the-art performance. We also empirically show that domain-specific class-relevant features are beneficial for domain generalization. The code is available at https://github.com/yuxi120407/INSURE.

Projector video compensation aims to cancel the geometric and photometric distortions caused by non-ideal projection surfaces and environments when projecting videos. Most existing projector compensation methods start by projecting and capturing a set of sampling images, followed by an offline compensation model training step. Thus, abundant user effort is required before the users can watch the video. Moreover, the sampling images have little prior knowledge of the video content and may lead to suboptimal results. To address these issues, this paper builds a video compensation system that can online adapt the compensation parameters. Our approach consists of five threads and can perform compensation, projection, capturing, and short-term and long-term model updates in parallel. Due to the parallel mechanism, rather than projecting and capturing hundreds of sampling images and training the model offline, we can directly use the projected and captured video frames for model updates on the fly. To quickly apply to the new environment, we introduce a deep learning-based compensation model that integrates a fixed transformer-based method and a novel CNN-based network. Moreover, for fast convergence and to reduce error accumulation during fine-tuning, we present a strategy that cooperates with short-term and long-term memory model updates. Experiments show that it significantly outperforms state-of-the-art baselines.

Enhancing model prediction confidence on target data is an important objective in Unsupervised Domain Adaptation (UDA). In this paper, we explore adversarial training on penultimate activations, i.e., input features of the final linear classification layer. We show that this strategy is more efficient and better correlated with the objective of boosting prediction confidence than adversarial training on input images or intermediate features, as used in previous works. Furthermore, with activation normalization commonly used in domain adaptation to reduce domain gap, we derive two variants and systematically analyze the effects of normalization on our adversarial training. This is illustrated both in theory and through empirical analysis on real adaptation tasks. Extensive experiments are conducted on popular UDA benchmarks under both standard setting and source-data free setting. The results validate that our method achieves the best scores against previous arts. Code is available at https://github.com/tsun/APA.

Due to the increasing computational demand of Deep Neural Networks (DNNs), companies and organizations have begun to outsource the training process. However, the externally trained DNNs can potentially be backdoor attacked. It is crucial to defend against such attacks, i.e., to postprocess a suspicious model so that its backdoor behavior is mitigated while its normal prediction power on clean inputs remain uncompromised. To remove the abnormal backdoor behavior, existing methods mostly rely on additional labeled clean samples. However, such requirement may be unrealistic as the training data are often unavailable to end users. In this paper, we investigate the possibility of circumventing such barrier. We propose a novel defense method that does not require training labels. Through a carefully designed layer-wise weight reinitialization and knowledge distillation, our method can effectively cleanse backdoor behaviors of a suspicious network with negligible compromise in its normal behavior. In experiments, we show that our method, trained without labels, is on-par with state-of-the-art defense methods trained using labels. We also observe promising defense results even on out-of-distribution data. This makes our method very practical. Code is available at: https://github.com/luluppang/BCU.

Deep learning-based edge detectors heavily rely on pixel-wise labels which are often provided by multiple annotators. Existing methods fuse multiple annotations using a simple voting process, ignoring the inherent ambiguity of edges and labeling bias of annotators. In this paper, we propose a novel uncertainty-aware edge detector (UAED), which employs uncertainty to investigate the subjectivity and ambiguity of diverse annotations. Specifically, we first convert the deterministic label space into a learnable Gaussian distribution, whose variance measures the degree of ambiguity among different annotations. Then we regard the learned variance as the estimated uncertainty of the predicted edge maps, and pixels with higher uncertainty are likely to be hard samples for edge detection. Therefore we design an adaptive weighting loss to emphasize the learning from those pixels with high uncertainty, which helps the network to gradually concentrate on the important pixels. UAED can be combined with various encoder-decoder backbones, and the extensive experiments demonstrate that UAED achieves superior performance consistently across multiple edge detection benchmarks. The source code is available at https://github.com/ZhouCX117/UAED.

Estimating the 6-DoF pose of a rigid object from a single RGB image is a crucial yet challenging task. Recent studies have shown the great potential of dense correspondence-based solutions, yet improvements are still needed to reach practical deployment. In this paper, we propose a novel pose estimation algorithm named CheckerPose, which improves on three main aspects. Firstly, CheckerPose densely samples 3D keypoints from the surface of the 3D object and finds their 2D correspondences progressively in the 2D image. Compared to previous solutions that conduct dense sampling in the image space, our strategy enables the correspondence searching in a 2D grid (i.e., pixel coordinate). Secondly, for our 3D-to-2D correspondence, we design a compact binary code representation for 2D image locations. This representation not only allows for progressive correspondence refinement but also converts the correspondence regression to a more efficient classification problem. Thirdly, we adopt a graph neural network to explicitly model the interactions among the sampled 3D keypoints, further boosting the reliability and accuracy of the correspondences. Together, these novel components make CheckerPose a strong pose estimation algorithm. When evaluated on the popular Linemod, Linemod-O, and YCB-V object pose estimation benchmarks, CheckerPose clearly boosts the accuracy of correspondence-based methods and achieves state-of-the-art performances. Code is available at https://github.com/RuyiLian/CheckerPose.

Active Domain Adaptation (ADA) queries the labels of a small number of selected target samples to help adapting a model from a source domain to a target domain. The local context of queried data is important, especially when the domain gap is large. However, this has not been fully explored by existing ADA works. In this paper, we propose a Local context-aware ADA framework, named LADA, to address this issue. To select informative target samples, we devise a novel criterion based on the local inconsistency of model predictions. Since the labeling budget is usually small, fine-tuning model on only queried data can be inefficient. We progressively augment labeled target data with the confident neighbors in a class-balanced manner. Experiments validate that the proposed criterion chooses more informative target samples than existing active selection strategies. Furthermore, our full method clearly surpasses recent ADA arts on various benchmarks. Code is available at https://github.com/tsun/LADA.

We observe that accurate and fast tracking in Internet of Things (IoT) devices is still a challenging problem. Several deep learning models have emerged which provide higher accuracy scores in object detection and tracking, however, due to their computationally expensive nature they are not useful in enabling real-time tracking at IoT devices. Correlation filters have emerged to show better speed in real-time tracking and provide good tracking results in cases of occlusion, rotation, illumination and other distractions. To get better speed as well as accuracy we use combination of correlation filter and deep learning methods. We propose a distributed tracking and verifying (DTAV) framework. Specifically, we run two object tracking algorithms, one on the client and another on the server. The algorithm run on the client is referred to as the Tracker, which is based on correlation filter and runs easily in real-time. The server hosts the verifier algorithm which performs high accuracy verification. Thus, while the client performs fast object tracking, the server's tracking algorithm verifies the output and corrects the server whenever required to maintain the accuracy of the model. We present our edge computing-based framework and discuss the motivation, system setup and series of experiments performed for the framework and present our experimental results. DTAV achieved 7.78% improvement on accuracy and 15% improvement in FPS.

Accurate tracking of transparent objects, such as glasses, plays a critical role in many robotic tasks such as robot-assisted living. Due to the adaptive and often reflective texture of such objects, traditional tracking algorithms that rely on general-purpose learned features suffer from reduced performance. Recent research has proposed to instill trans-parency awareness into existing general object trackers by fusing purpose-built features. However, with the existing fusion techniques, the addition of new features causes a change in the latent space making it impossible to incorporate transparency awareness on trackers with fixed latent spaces. For example, many of the current days' transformer-based trackers are fully pre-trained and are sensitive to any latent space perturbations. In this paper, we present a new feature fusion technique that integrates transparency information into a fixed feature space, enabling its use in a broader range of trackers. Our proposed fusion module, composed of a transformer encoder and an MLP module, leverages key query-based transformations to embed the transparency information into the tracking pipeline. We also present a new two-step training strategy for our fusion module to effectively merge transparency features. We propose a new tracker architecture that uses our fusion techniques to achieve superior results for transparent object tracking. Our proposed method achieves competitive results with state-of-the-art trackers on TOTB, which is the largest transparent object tracking benchmark recently released. Our results and the implementation of code will be made publicly available at https://github.com/kalyan0510/TOTEM.

Color structured light (SL) plays an important role in spatial augmented reality and shape reconstruction. Compared to traditional non-color multi-shot SL, it has the advantage of fewer projections, and can even achieve single-shot. However, distortions caused by ambient light and imaging devices limit color SL’s applicability and accuracy. A common solution is to apply color adaptation techniques to cancel the disturbances. Previous studies focus on either robust fixed color patterns or adaptation approaches that may require preliminary geometric calibrations. In this paper, we propose an approach that can efficiently adapt color SL to arbitrary ambient light and imaging devices’ color responses, without device response function calibration or geometric calibration. First, we design a novel algorithm to quickly find the most distinct colors that are easily separable under a new environment and device setup. Then, we design a maximum a posteriori (MAP)-based color detection algorithm that can utilize ambient light and device priors to robustly detect the SL colors. In experiments, our adaptive color SL outperforms previous methods in both calibration and shape reconstruction tasks across a variety of setups.

Full projector compensation is a practical task of projector-camera systems. It aims to find a projector input image, named compensation image, such that when projected it cancels the geometric and photometric distortions due to the physical environment and hardware. State-of-the-art methods use deep learning to address this problem and show promising performance for low-resolution setups. However, directly applying deep learning to high-resolution setups is impractical due to the long training time and high memory cost. To address this issue, this paper proposes a practical full compensation solution. Firstly, we design an attention-based grid refinement network to improve geometric correction quality. Secondly, we integrate a novel sampling scheme into an end-to-end compensation network to alleviate computation and introduce attention blocks to preserve key features. Finally, we construct a benchmark dataset for high-resolution projector full compensation. In experiments, our method demonstrates clear advantages in both efficiency and quality.

Tracking multiple objects in videos relies on modeling the spatial-temporal interactions of the objects. In this paper, we propose TransMOT, which leverages powerful graph transformers to efficiently model the spatial and temporal interactions among the objects. TransMOT is capable of effectively modeling the interactions of a large number of objects by arranging the trajectories of the tracked targets and detection candidates as a set of sparse weighted graphs, and constructing a spatial graph transformer encoder layer, a temporal transformer encoder layer, and a spatial graph transformer decoder layer based on the graphs. Through end-to-end learning, TransMOT can exploit the spatial-temporal clues to directly estimate association from a large number of loosely filtered detection predictions for robust MOT in complex scenes. The proposed method is evaluated on multiple benchmark datasets, including MOT15, MOT16, MOT17, and MOT20, and it achieves state-of-the-art performance on all the datasets.

Subspace clustering is a classical technique that has been widely used for human motion segmentation and other related tasks. However, existing segmentation methods often cluster data without guidance from prior knowledge, resulting in unsatisfactory segmentation results. To this end, we propose a novel Consistency and Diversity induced human Motion Segmentation (CDMS) algorithm. Specifically, our model factorizes the source and target data into distinct multi-layer feature spaces, in which transfer subspace learning is conducted on different layers to capture multi-level information. A multi-mutual consistency learning strategy is carried out to reduce the domain gap between the source and target data. In this way, the domain-specific knowledge and domain-invariant properties can be explored simultaneously. Besides, a novel constraint based on the Hilbert Schmidt Independence Criterion (HSIC) is introduced to ensure the diversity of multi-level subspace representations, which enables the complementarity of multi-level representations to be explored to boost the transfer learning performance. Moreover, to preserve the temporal correlations, an enhanced graph regularizer is imposed on the learned representation coefficients and the multi-level representations of the source data. The proposed model can be efficiently solved using the Alternating Direction Method of Multipliers (ADMM) algorithm. Extensive experimental results on public human motion datasets demonstrate the effectiveness of our method against several state-of-the-art approaches.

Differentiable ARchiTecture Search, i.e., DARTS, has drawn great attention in neural architecture search. It tries to find the optimal architecture in a shallow search network and then measures its performance in a deep evaluation network. The independent optimization of the search and evaluation networks, however, leaves a room for potential improvement by allowing interaction between the two networks. To address the problematic optimization issue, we propose new joint optimization objectives and a novel Cyclic Differentiable ARchiTecture Search framework, dubbed CDARTS. Considering the structure difference, CDARTS builds a cyclic feedback mechanism between the search and evaluation networks with introspective distillation. First, the search network generates an initial architecture for evaluation, and the weights of the evaluation network are optimized. Second, the architecture weights in the search network are further optimized by the label supervision in classification, as well as the regularization from the evaluation network through feature distillation. Repeating the above cycle results in a joint optimization of the search and evaluation networks and thus enables the evolution of the architecture to fit the final evaluation network. The experiments and analysis on CIFAR, ImageNet and NATS-Bench [95] demonstrate the effectiveness of the proposed approach over the state-of-the-art ones. Specifically, in the DARTS search space, we achieve 97.52% top-1 accuracy on CIFAR10 and 76.3% top-1 accuracy on ImageNet. In the chain-structured search space, we achieve 78.2% top-1 accuracy on ImageNet, which is 1.1% higher than EfficientNet-B0. Our code and models are publicly available at https://github.com/microsoft/Cream.

Salient object detection (SOD) in optical remote sensing images (RSIs), or RSI-SOD, is an emerging topic in understanding optical RSIs. However, due to the difference between optical RSIs and natural scene images (NSIs), directly applying NSI-SOD methods to optical RSIs fails to achieve satisfactory results. In this article, we propose a novel adjacent context coordination network (ACCoNet) to explore the coordination of adjacent features in an encoder-decoder architecture for RSI-SOD. Specifically, ACCoNet consists of three parts: 1) an encoder; 2) adjacent context coordination modules (ACCoMs); and 3) a decoder. As the key component of ACCoNet, ACCoM activates the salient regions of output features of the encoder and transmits them to the decoder. ACCoM contains a local branch and two adjacent branches to coordinate the multilevel features simultaneously. The local branch highlights the salient regions in an adaptive way, while the adjacent branches introduce global information of adjacent levels to enhance salient regions. In addition, to extend the capabilities of the classic decoder block (i.e., several cascaded convolutional layers), we extend it with two bifurcations and propose a bifurcation-aggregation block (BAB) to capture the contextual information in the decoder. Extensive experiments on two benchmark datasets demonstrate that the proposed ACCoNet outperforms 22 state-of-the-art methods under nine evaluation metrics, and runs up to 81 fps on a single NVIDIA Titan X GPU. The code and results of our method are available at https://github.com/MathLee/ACCoNet.

Existing methods for Salient Object Detection in Optical Remote Sensing Images (ORSI-SOD) mainly adopt Convolutional Neural Networks (CNNs) as the backbone, such as VGG and ResNet. Since CNNs can only extract features within certain receptive fields, most ORSI-SOD methods generally follow the local-to-contextual paradigm. In this paper, we propose a novel Global Extraction Local Exploration Network (GeleNet) for ORSI-SOD following the global-to-local paradigm. Specifically, GeleNet first adopts a transformer backbone to generate four-level feature embeddings with global long-range dependencies. Then, GeleNet employs a Direction-aware Shuffle Weighted Spatial Attention Module (D-SWSAM) and its simplified version (SWSAM) to enhance local interactions, and a Knowledge Transfer Module (KTM) to further enhance cross-level contextual interactions. D-SWSAM comprehensively perceives the orientation information in the lowest-level features through directional convolutions to adapt to various orientations of salient objects in ORSIs, and effectively enhances the details of salient objects with an improved attention mechanism. SWSAM discards the direction-aware part of D-SWSAM to focus on localizing salient objects in the highest-level features. KTM models the contextual correlation knowledge of two middle-level features of different scales based on the self-attention mechanism, and transfers the knowledge to the raw features to generate more discriminative features. Finally, a saliency predictor is used to generate the saliency map based on the outputs of the above three modules. Extensive experiments on three public datasets demonstrate that the proposed GeleNet outperforms relevant state-of-the-art methods. The code and results of our method are available at https://github.com/MathLee/GeleNet.

Image fusion synthesizes a new image from multiple images of the same scene. The synthesized image should be suitable for human visual perception and follow-up high-level image-processing tasks. However, existing methods focus on fusing low-level features, ignoring high-level semantic perception information. We propose a new end-to-end model to obtain a more semantically consistent image in infrared and visible image fusion, termedsemantic-supervised dual-discriminator generative adversarial network(SDDGAN). In particular, we design an information quantity discrimination (IQD) block to guide fusion progress. For each source image, the block determines the weight for preserving each semantic object’s feature. By this way, the generator learns to fuse various semantic objects via different weights to preserve their characteristics. Moreover, the dual discriminator is employed to identify the distribution of infrared and visible information in the fused image. Each discriminator acts on a certain modality (infrared/visible) of different semantic objects in the fused image to preserve and enhance their modality features. Thus, our fused image is more informative. Both the thermal radiation in the infrared image and the visible image texture details can be well preserved. Qualitative and quantitative experiments demonstrate the superiority of our SDDGAN over state-of-the-art methods in terms of visual effects, efficiency, and quantitative metrics.

In this paper, we present ARCHIE++, a testing framework for conducting AR system testing and collecting user feedback in the wild. Our system addresses challenges in AR testing practices by aggregating usability feedback data (collected in situ) with system performance data from that same time period. These data packets can then be leveraged to identify edge cases encountered by testers during unconstrained usage scenarios. We begin by presenting a set of current trends in performing human testing of AR systems, identified by reviewing a selection of recent work from leading conferences in mixed reality, human factors, and mobile and pervasive systems. From the trends, we identify a set of challenges to be faced when attempting to adopt these practices to testing in the wild. These challenges are used to inform the design of our framework, which provides a cloud-enabled and device-agnostic way for AR systems developers to improve their knowledge of environmental conditions and to support scalability and reproducibility when testing in the wild. We then present a series of case studies demonstrating how ARCHIE++ can be used to support a range of AR testing scenarios, and demonstrate the limited overhead of the framework through a series of evaluations. We close with additional discussion on the design and utility of ARCHIE++ under various edge conditions.

In this work, we focus on the label layout problem: specifying the positions of overlaid virtual annotations in Virtual/Augmented Reality scenarios. Designing a layout of labels that does not violate domain-specific design requirements, while at the same time satisfying aesthetic and functional principles of good design, can be a daunting task even for skilled visual designers. Presenting the annotations in 3D object space instead of projection space, allows for the preservation of spatial and depth cues. This results in stable layouts in dynamic environments, since the annotations are anchored in 3D space. In this paper we make two major contributions. First, we propose a technique for managing the layout and rendering of annotations in Virtual/Augmented Reality scenarios by manipulating the annotations directly in 3D space. For this, we make use of Artificial Potential Fields and use 3D geometric constraints to adapt them in 3D space. Second, we introduce PartLabeling: an open source platform in the form of a web application that acts as a much-needed generic framework allowing to easily add labeling algorithms and 3D models. This serves as a catalyst for researchers in this field to make their algorithms and implementations publicly available, as well as ensure research reproducibility. The PartLabeling framework relies on a dataset that we generate as a subset of the original PartNet dataset [17] consisting of models suitable for the label management task. The dataset consists of 1,000 3D models with part annotations.

Graph neural networks are a promising architecture for learning and inference with graph-structured data. Yet, how to generate informative, fixed dimensional features for graphs with varying size and topology can still be challenging. Typically, this is achieved through graph-pooling, which summarizes a graph by compressing all its nodes into a single vector. Is such a “collapsing-style” graph-pooling the only choice for graph classification? From complex system’s point of view, properties of a complex system arise largely from the interaction among its components. Therefore, we speculate that preserving the interacting relation between parts, instead of pooling them together, could benefit system level prediction. To verify this, we propose SLIM, a graph neural network model for Structural Landmarking and Interaction Modelling. The main idea is to compute a set of end-to-end optimizable sub-structure landmarks, so that any input graph can be projected onto these (spatially) local structural representatives for a faithful, global characterization. By doing so, explicit interaction between component parts of a graph can be leveraged directly in generating discriminative graph representation. Encouraging results are observed on benchmark datasets for graph classification, demonstrating the value of interaction modelling in the design of graph neural networks.

We propose HandLer, a novel convolutional architecture that can jointly detect and track hands online in unconstrained videos. HandLer is based on Cascade-RCNN with additional three novel stages. The first stage is Forward Propagation, where the features from frame t −1 are propagated to frame t based on previously detected hands and their estimated motion. The second stage is the Detection and Backward Regression, which uses outputs from the forward propagation to detect hands for frame t and their relative offset in frame t −1. The third stage uses an off-the-shelf human pose method to link any fragmented hand tracklets. We train the forward propagation and backward regression and detection stages end-to-end together with the other Cascade-RCNN components. To train and evaluate HandLer, we also contribute YouTube-Hand, the first challenging large-scale dataset of unconstrained videos annotated with hand locations and their trajectories. Experiments on this dataset and other benchmarks show that HandLer outperforms the existing state-of-the-art tracking algorithms by a large margin. Code and data are available at https://vision.cs.stonybrook.edu/~mingzhen/handler/.

Convolutional neural networks have made significant progresses in edge detection by progressively exploring the context and semantic features. However, local details are gradually suppressed with the enlarging of receptive fields. Recently, vision transformer has shown excellent capability in capturing long-range dependencies. Inspired by this, we propose a novel transformer-based edge detector, Edge Detection TransformER (EDTER), to extract clear and crisp object boundaries and meaningful edges by exploiting the full image context information and detailed local cues simultaneously. EDTER works in two stages. In Stage I, a global transformer encoder is used to capture long-range global context on coarse-grained image patches. Then in Stage II, a local transformer encoder works on fine-grained patches to excavate the short-range local cues. Each transformer encoder is followed by an elaborately designed Bi-directional Multi-Level Aggregation decoder to achieve high-resolution features. Finally, the global context and local cues are combined by a Feature Fusion Module and fed into a decision head for edge prediction. Extensive experiments on BSDS500, NYUDv2, and Multicue demonstrate the superiority of EDTER in comparison with state-of-the-arts. The source code is available at https://github.com/MengyangPu/EDTER.

Optical coherence tomography angiography (OCTA) is an important imaging modality in many bioengineering tasks. The image quality of OCTA, however, is often degraded by Bulk Motion Artifacts (BMA), which are due to micromotion of subjects and typically appear as bright stripes surrounded by blurred areas. State-of-the-art methods usually treat BMA removal as a learning-based image inpainting problem, but require numerous training samples with nontrivial annotation. In addition, these methods discard the rich structural and appearance information carried in the BMA stripe region. To address these issues, in this paper we propose a self-supervised content-aware BMA removal model. First, the gradient-based structural information and appearance feature are extracted from the BMA area and injected into the model to capture more connectivity. Second, with easily collected defective masks, the model is trained in a self-supervised manner, in which only the clear areas are used for training while the BMA areas for inference. With the structural information and appearance feature from noisy image as references, our model can remove larger BMA and produce better visualizing result. In addition, only 2D images with defective masks are involved, hence improving the efficiency of our method. Experiments on OCTA of mouse cortex demonstrate that our model can remove most BMA with extremely large sizes and inconsistent intensities while previous methods fail.

Unsupervised Domain Adaptation (UDA) aims to leverage a label-rich source domain to solve tasks on a related unlabeled target domain. It is a challenging problem especially when a large domain gap lies between the source and target domains. In this paper we propose a novel solution named SSRT (Safe Self-Refinement for Transformer-based domain adaptation), which brings improvement from two aspects. First, encouraged by the success of vision transformers in various vision tasks, we arm SSRT with a transformer backbone. We find that the combination of vision transformer with simple adversarial adaptation surpasses best reported Convolutional Neural Network (CNN)-based results on the challenging DomainNet benchmark, showing its strong transferable feature representation. Second, to reduce the risk of model collapse and improve the effectiveness of knowledge transfer between domains with large gaps, we propose a Safe Self-Refinement strategy. Specifically, SSRT utilizes predictions of perturbed target domain data to refine the model. Since the model capacity of vision transformer is large and predictions in such challenging tasks can be noisy, a safe training mechanism is designed to adaptively adjust learning configuration. Extensive evaluations are conducted on several widely tested UDA benchmarks and SSRT achieves consistently the best performances, including 85.43% on Office-Home, 88.76% on VisDA-2017 and 45.2% on DomainNet.

Recently Transformer has been largely explored in tracking and shown state-of-the-art (SOTA) performance. However, existing efforts mainly focus on fusing and enhancing features generated by convolutional neural networks (CNNs). The potential of Transformer in representation learning remains under-explored. In this paper, we aim to further unleash the power of Transformer by proposing a simple yet efficient fully-attentional tracker, dubbed SwinTrack, within classic Siamese framework. In particular, both representation learning and feature fusion in SwinTrack leverage the Transformer architecture, enabling better feature interactions for tracking than pure CNN or hybrid CNN-Transformer frameworks. Besides, to further enhance robustness, we present a novel motion token that embeds historical target trajectory to improve tracking by providing temporal context. Our motion token is lightweight with negligible computation but brings clear gains. In our thorough experiments, SwinTrack exceeds existing approaches on multiple benchmarks. Particularly, on the challenging LaSOT, SwinTrack sets a new record with 0.713 SUC score. It also achieves SOTA results on other benchmarks. We expect SwinTrack to serve as a solid baseline for Transformer tracking and facilitate future research. Our codes and results are released at https://github.com/LitingLin/SwinTrack.

Light-based adversarial attacks use spatial augmented reality (SAR) techniques to fool image classifiers by altering the physical light condition with a controllable light source, e.g., a projector. Compared with physical attacks that place hand-crafted adversarial objects, projector-based ones obviate modifying the physical entities, and can be performed transiently and dynamically by altering the projection pattern. However, subtle light perturbations are insufficient to fool image classifiers, due to the complex environment and project-and-capture process. Thus, existing approaches focus on projecting clearly perceptible adversarial patterns, while the more interesting yet challenging goal, stealthy projector-based attack, remains open. In this paper, for the first time, we formulate this problem as an end-to-end differentiable process and propose a Stealthy Projector-based Adversarial Attack (SPAA) solution. In SPAA, we approximate the real Project-and-Capture process using a deep neural network named PCNet, then we include PCNet in the optimization of projector-based attacks such that the generated adversarial projection is physically plausible. Finally, to generate both robust and stealthy adversarial projections, we propose an algorithm that uses minimum perturbation and adversarial confidence thresholds to alternate between the adversarial loss and stealthiness loss optimization. Our experimental evaluations show that SPAA clearly outperforms other methods by achieving higher attack success rates and meanwhile being stealthier, for both targeted and untargeted attacks.

In this work, we consider the problem of self-supervised Moving Object Detection (MOD) in video, where no ground truth is involved in both training and inference phases. Recently, an adversarial learning framework is proposed [32] to leverage inherent temporal information for MOD. While showing great promising results, it uses single scale temporal information and may meet problems when dealing with a deformable object under multi-scale motion in different parts. Additional challenges can arise from the moving camera, which results in the failure of the motion independence hypothesis and locally independent background motion. To deal with these problems, we propose a Multimotion and Appearance Self-supervised Network (MASNet) to introduce multi-scale motion information and appearance information of scene for MOD. In particular, a moving object, especially the deformable, usually consists of moving regions at various temporal scales. Introducing multiscale motion can aggregate these regions to form a more complete detection. Appearance information can serve as another cue for MOD when the motion independence is not reliable and for removing false detection in background caused by locally independent background motion. To encode multi-scale motion and appearance, in MASNet we respectively design a multi-branch flow encoding module and an image inpainter module. The proposed modules and MASNet are extensively evaluated on the DAVIS dataset to demonstrate the effectiveness and superiority to state-of-the-art self-supervised methods.

Geometric model fitting is a fundamental task in computer vision, which serves as the pre-requisite of many downstream applications. While the problem has a simple intrinsic structure where the solution can be parameterized within a few degrees of freedom, the ubiquitously existing outliers are the main challenge. In previous studies, random sampling techniques have been established as the practical choice, since optimization-based methods are usually too time-demanding. This prospective study is intended to design efficient algorithms that benefit from a general optimization-based view. In particular, two important types of loss functions are discussed, \emph{i.e.} truncated and$l_1$losses, and efficient solvers have been derived for both upon specific approximations. Based on this philosophy, a class of algorithms are introduced to perform deterministic search for the inliers or geometric model. Recommendations are made based on theoretical and experimental analyses. Compared with the existing solutions, the proposed methods are both simple in computation and robust to outliers. Extensive experiments are conducted on publicly available datasets for geometric estimation, which demonstrate the superiority of our methods compared with the state-of-the-art ones. Additionally, we apply our method to the recent benchmark for wide-baseline stereo evaluation, leading to a significant improvement of performance.

Real-time semantic video segmentation is a challenging task due to the strict requirements of inference speed. Recent approaches mainly devote great efforts to reducing the model size for high efficiency. In this paper, we rethink this problem from a different viewpoint: using knowledge contained in compressed videos. We propose a simple and effective framework, dubbed TapLab, to tap into resources from the compressed domain. Specifically, we design a fast feature warping module using motion vectors for acceleration. To reduce the noise introduced by motion vectors, we design a residual-guided correction module and a residual-guided frame selection module using residuals. TapLab significantly reduces redundant computations of the state-of-the-art fast semantic image segmentation models, running 3 to 10 times faster with controllable accuracy degradation. The experimental results show that TapLab achieves 70.6 percent mIoU on the Cityscapes dataset at 99.8 FPS with a single GPU card for the 1024×2048 videos. A high-speed version even reaches the speed of 160+ FPS. Code will be available soon at https://github.com/Sixkplus/TapLab.

Full projector compensation aims to modify a projector input image to compensate for both geometric and photometric disturbance of the projection surface. Traditional methods usually solve the two parts separately and may suffer from suboptimal solutions. In this paper, we propose the first end-to-end differentiable solution, named CompenNeSt++, to solve the two problems jointly. First, we propose a novel geometric correction subnet, named WarpingNet, which is designed with a cascaded coarse-to-fine structure to learn the sampling grid directly from sampling images. Second, we propose a novel photometric compensation subnet, named CompenNeSt, which is designed with a siamese architecture to capture the photometric interactions between the projection surface and the projected images, and to use such information to compensate the geometrically corrected images. By concatenating WarpingNet with CompenNeSt, CompenNeSt++ accomplishes full projector compensation and is end-to-end trainable. Third, to improve practicability, we propose a novel synthetic data-based pre-training strategy to significantly reduce the number of training images and training time. Moreover, we construct the first setup-independent full compensation benchmark to facilitate future studies. In thorough experiments, our method shows clear advantages over prior art with promising compensation quality and meanwhile being practically convenient.

As an essential problem in computer vision, salient object detection (SOD) has attracted an increasing amount of research attention over the years. Recent advances in SOD are predominantly led by deep learning-based solutions (named deep SOD). To enable in-depth understanding of deep SOD, in this paper, we provide a comprehensive survey covering various aspects, ranging from algorithm taxonomy to unsolved issues. In particular, we first review deep SOD algorithms from different perspectives, including network architecture, level of supervision, learning paradigm, and object-/instance-level detection. Following that, we summarize and analyze existing SOD datasets and evaluation metrics. Then, we benchmark a large group of representative SOD models, and provide detailed analyses of the comparison results. Moreover, we study the performance of SOD algorithms under different attribute settings, which has not been thoroughly explored previously, by constructing a novel SOD dataset with rich attribute annotations covering various salient object types, challenging factors, and scene categories. We further analyze, for the first time in the field, the robustness of SOD models to random input perturbations and adversarial attacks. We also look into the generalization and difficulty of existing SOD datasets. Finally, we discuss several open issues of SOD and outline future research directions. All the saliency prediction maps, our constructed dataset with annotations, and codes for evaluation are publicly available at https://github.com/wenguanwang/SODsurvey.

This study proposes a novel unified and unsupervised end-to-end image fusion network, termed as U2Fusion, which is capable of solving different fusion problems, including multi-modal, multi-exposure, and multi-focus cases. Using feature extraction and information measurement, U2Fusion automatically estimates the importance of corresponding source images and comes up with adaptive information preservation degrees. Hence, different fusion tasks are unified in the same framework. Based on the adaptive degrees, a network is trained to preserve the adaptive similarity between the fusion result and source images. Therefore, the stumbling blocks in applying deep learning for image fusion, e.g., the requirement of ground-truth and specifically designed metrics, are greatly mitigated. By avoiding the loss of previous fusion capabilities when training a single model for different tasks sequentially, we obtain a unified model that is applicable to multiple fusion tasks. Moreover, a new aligned infrared and visible image dataset, RoadScene (available at https://github.com/hanna-xu/RoadScene), is released to provide a new option for benchmark evaluation. Qualitative and quantitative experimental results on three typical image fusion tasks validate the effectiveness and universality of U2Fusion. Our code is publicly available at https://github.com/hanna-xu/U2Fusion.

Drones, or general UAVs, equipped with cameras have been fast deployed with a wide range of applications, including agriculture, aerial photography, and surveillance. Consequently, automatic understanding of visual data collected from drones becomes highly demanding, bringing computer vision and drones more and more closely. To promote and track the developments of object detection and tracking algorithms, we have organized three challenge workshops in conjunction with ECCV 2018, ICCV 2019 and ECCV 2020, attracting more than 100 teams around the world. We provide a large-scale drone captured dataset, VisDrone, which includes four tracks, i.e., (1) image object detection, (2) video object detection, (3) single object tracking, and (4) multi-object tracking. In this paper, we first present a thorough review of object detection and tracking datasets and benchmarks, and discuss the challenges of collecting large-scale drone-based object detection and tracking datasets with fully manual annotations. After that, we describe our VisDrone dataset, which is captured over various urban/suburban areas of 14 different cities across China from North to South. Being the largest such dataset ever published, VisDrone enables extensive evaluation and investigation of visual analysis algorithms for the drone platform. We provide a detailed analysis of the current state of the field of large-scale object detection and tracking on drones, and conclude the challenge as well as propose future directions. We expect the benchmark largely boost the research and development in video analysis on drone platforms. All the datasets and experimental results can be downloaded from https://github.com/VisDrone/VisDrone-Dataset.

Salient object detection in optical remote sensing images (ORSI-SOD) has been widely explored for understanding ORSIs. However, previous methods focus mainly on improving the detection accuracy while neglecting the cost in memory and computation, which may hinder their real-world applications. In this article, we propose a novel lightweight ORSI-SOD solution, named CorrNet, to address these issues. In CorrNet, we first lighten the backbone (VGG-16) and build a lightweight subnet for feature extraction. Then, following the coarse-to-fine strategy, we generate an initial coarse saliency map from high-level semantic features in a correlation module (CorrM). The coarse saliency map serves as the location guidance for low-level features. In CorrM, we mine the object location information between high-level semantic features through the cross-layer correlation operation. Finally, based on low-level detailed features, we refine the coarse saliency map in the refinement subnet equipped with dense lightweight refinement blocks (DLRBs) and produce the final fine saliency map. By reducing the parameters and computations of each component, CorrNet ends up having only 4.09M parameters and running with 21.09G FLOPs. Experimental results on two public datasets demonstrate that our lightweight CorrNet achieves competitive or even better performance compared with 26 state-of-the-art methods (including 16 large CNN-based methods and two lightweight methods), and meanwhile enjoys the clear memory and run-time efficiency. The code and results of our method are available athttps://github.com/MathLee/CorrNet.

In the computer vision community, great progresses have been achieved in salient object detection from natural scene images (NSI-SOD); by contrast, salient object detection in optical remote sensing images (RSI-SOD) remains to be a challenging emerging topic. The unique characteristics of optical RSIs, such as scales, illuminations, and imaging orientations, bring significant differences between NSI-SOD and RSI-SOD. In this article, we propose a novel multi-content complementation network (MCCNet) to explore the complementarity of multiple content for RSI-SOD. Specifically, MCCNet is based on the general encoder–decoder architecture, and contains a novel key component named multi-content complementation module (MCCM), which bridges the encoder and the decoder. In MCCM, we consider multiple types of features that are critical to RSI-SOD, including foreground features, edge features, background features, and global image-level features, and exploit the content complementarity between them to highlight salient regions over various scales in RSI features through the attention mechanism. Besides, we comprehensively introduce pixel-level, map-level, and metric-aware losses in the training phase. Extensive experiments on two popular datasets demonstrate that the proposed MCCNet outperforms 23 state-of-the-art methods, including both NSI-SOD and RSI-SOD methods. The code and results of our method are available athttps://github.com/MathLee/MCCNet.

top-performing object detectors depend heavily on backbone networks, whose advances bring consistent performance gains through exploring more effective network structures. In this paper, we propose a novel and flexible backbone framework, namely CBNet, to construct high-performance detectors using existing open-source pre-trained backbones under the pre-training fine-tuning paradigm. In particular, CBNet architecture groups multiple identical backbones, which are connected through composite connections. Specifically, it integrates the high- and low-level features of multiple identical backbone networks and gradually expands the receptive field to more effectively perform object detection. We also propose a better training strategy with auxiliary supervision for CBNet-based detectors. CBNet has strong generalization capabilities for different backbones and head designs of the detector architecture. Without additional pre-training of the composite backbone, CBNet can be adapted to various backbones (i.e., CNN-based vs. Transformer-based) and head designs of most mainstream detectors (i.e., one-stage vs. two-stage, anchor-based vs. anchor-free-based). Experiments provide strong evidence that, compared with simply increasing the depth and width of the network, CBNet introduces a more efficient, effective, and resource-friendly way to build high-performance backbone networks. Particularly, our CB-Swin-L achieves 59.4% box AP and 51.6% mask AP on COCO test-dev under the single-model and single-scale testing protocol, which are significantly better than the state-of-the-art results (i.e., 57.7% box AP and 50.2% mask AP) achieved by Swin-L, while reducing the training time by 6×. With multi-scale testing, we push the current best single model result to a new record of 60.1% box AP and 52.3% mask AP without using extra training data. Code is available at https://github.com/VDIGPKU/CBNetV2.

Mobile augmented reality systems are becoming increasingly common and powerful, with applications in such domains as healthcare, manufacturing, education, and more. This rise in popularity is thanks in part to the functionalities offered by commercially available vision libraries such as ARCore, Vuforia, and Google’s ML Kit; however, these libraries also give rise to the possibility of a hidden operations threat , that is, the ability of a malicious or incompetent application developer to conduct additional vision operations behind the scenes of an otherwise honest AR application without alerting the end-user. In this article, we present the privacy risks associated with the hidden operations threat and propose a framework for application development and runtime permissions targeted specifically at preventing the execution of hidden operations. We follow this with a set of experimental results, exploring the feasibility and utility of our system in differentiating between user-expectation-compliant and non-compliant AR applications during runtime testing, for which preliminary results demonstrate accuracy of up to 71%. We conclude with a discussion of open problems in the areas of software testing and privacy standards in mobile AR systems.

Due to recent advances in learning-based semantic segmentation, road scene parsing can usually achieve satisfactory results under normal illumination conditions. However, training a robust model for parsing nighttime road scenes is still very challenging, especially when semantic labels of training samples are absent. In this paper, we propose a convolutional neural network (CNN)-based method for parsing nighttime road scenes in an unsupervised manner. The proposed system includes an appearance transferring module and a segmentation module, which are coupled together and learned in an end-to-end fashion. The appearance transferring module aims to transfer unlabeled images acquired during both daytime and nighttime into a shared latent feature space that encodes the image content of both scenes at the semantic level. Then, the segmentation module is used to map the feature to its corresponding semantic labels. To better evaluate the proposed model, we also construct a new semantic segmentation dataset including 1,566 nighttime images. The extensive experimental results on the proposed benchmark illustrate that the proposed model achieves significant improvement compared with the baselines as well as a recently released system.

Face detection from low-light images is challenging due to limited photons and inevitable noise, which, to make the task even harder, are often spatially unevenly distributed. A natural solution is to borrow the idea frommulti-exposure, which captures multiple shots to obtain well-exposed images under challenging conditions. High-quality implementation/approximation of multi-exposure from a single image is however nontrivial. Fortunately, as shown in this paper, neither is such high-quality necessary since our task isface detectionrather thanimage enhancement. Specifically, we propose a novelRecurrent Exposure Generation (REG)module and couple it seamlessly with aMulti-Exposure Detection (MED)module, and thus significantly improve face detection performance by effectively inhibiting non-uniform illumination and noise issues. REG produces progressively and efficiently intermediate images corresponding to various exposure settings, and such pseudo-exposures are then fused by MED to detect faces across different lighting conditions. The proposed method, namedREGDet, is the first ‘detection-with-enhancement’ framework for low-light face detection. It not only encourages rich interaction and feature fusion across different illumination levels, but also enables effective end-to-end learning of the REG component to be better tailored for face detection. Moreover, as clearly shown in our experiments, REG can be flexibly coupled with different face detectors without extra low/normal-light image pairs for training. We tested REGDet on the DARK FACE low-light face benchmark with thorough ablation study, where REGDet outperforms previous state-of-the-arts by a significant margin, with only negligible extra parameters.

Multiple Object Tracking (MOT) has witnessed remarkable advances in recent years. However, existing studies dominantly request prior knowledge of the tracking target (eg, pedestrians), and hence may not generalize well to unseen categories. In contrast, Generic Multiple Object Tracking (GMOT), which requires little prior information about the target, is largely under-explored. In this paper, we make contributions to boost the study of GMOT in three aspects. First, we construct the first publicly available dense GMOT dataset, dubbed GMOT-40, which contains 40 carefully annotated sequences evenly distributed among 10 object categories. In addition, two tracking protocols are adopted to evaluate different characteristics of tracking algorithms. Second, by noting the lack of devoted tracking algorithms, we have designed a series of baseline GMOT algorithms. Third, we perform a thorough evaluations on GMOT-40, involving popular MOT algorithms (with necessary modifications) and the proposed baselines. The GMOT-40 benchmark is publicly available at https://github.com/Spritea/GMOT40.

Despite remarkable progress achieved, most neural architecture search (NAS) methods focus on searching for one single accurate and robust architecture. To further build models with better generalization capability and performance, model ensemble is usually adopted and performs better than stand-alone models. Inspired by the merits of model ensemble, we propose to search for multiple diverse models simultaneously as an alternative way to find powerful models. Searching for ensembles is non-trivial and has two key challenges: enlarged search space and potentially more complexity for the searched model. In this paper, we propose a one-shot neural ensemble architecture search (NEAS) solution that addresses the two challenges. For the first challenge, we introduce a novel diversity-based metric to guide search space shrinking, considering both the potentiality and diversity of candidate operators. For the second challenge, we enable a new search dimension to learn layer sharing among different models for efficiency purposes. The experiments on ImageNet clearly demonstrate that our solution can improve the supernet’s capacity of ranking ensemble architectures, and further lead to better search results. The discovered architectures achieve superior performance compared with state-of-the-arts such as MobileNetV3 and EfficientNet families under aligned settings. Moreover, we evaluate the generalization ability and robustness of our searched architecture on the COCO detection benchmark and achieve a 3.1% improvement on AP compared with MobileNetV3. Codes and models are available here.

Recently, neural architecture search (NAS) has been exploited to design feature pyramid networks (FPNs) and achieved promising results for visual object detection. Encouraged by the success, we propose a novel One-Shot Path Aggregation Network Architecture Search (OPANAS) algorithm, which significantly improves both searching efficiency and detection accuracy. Specifically, we first introduce six heterogeneous information paths to build our search space, namely top-down, bottom-up, fusing-splitting, scale-equalizing, skip-connect and none. Second, we propose a novel search space of FPNs, in which each FPN candidate is represented by a densely-connected directed acyclic graph (each node is a feature pyramid and each edge is one of the six heterogeneous information paths). Third, we propose an efficient one-shot search method to find the optimal path aggregation architecture; specifically, we first train a super-net and then find the optimal candidate with an evolutionary algorithm. Experimental results demonstrate the efficacy of the proposed OPANAS for object detection: (1) OPANAS is more efficient than state-of-the-art methods (e.g., NAS-FPN and Auto-FPN) at significantly smaller searching cost (e.g., only 4 GPU days on MS-COCO); (2) the optimal architecture found by OPANAS significantly improves main-stream detectors including RetinaNet, Faster R-CNN and Cascade R-CNN, by 2.3∼3.2 % mAP compared to their FPN counterparts; and (3) a new state-of-the-art accuracy-speed trade-off (52.2 % mAP at 7.6 FPS) is achieved at smaller training costs than comparable recent arts. Code will be released at https://github.com/VDIGPKU/OPANAS.

Recently, pure transformer-based models have shown great potentials for vision tasks such as image classification and detection. However, the design of transformer networks is challenging. It has been observed that the depth, embedding dimension, and number of heads can largely affect the performance of vision transformers. Previous models configure these dimensions based upon manual crafting. In this work, we propose a new one-shot architecture search framework, namely AutoFormer, dedicated to vision transformer search. AutoFormer entangles the weights of different blocks in the same layers during supernet training. Benefiting from the strategy, the trained supernet allows thousands of subnets to be very well-trained. Specifically, the performance of these subnets with weights inherited from the supernet is comparable to those retrained from scratch. Besides, the searched models, which we refer to AutoFormers, surpass the recent state-of-the-arts such as ViT and DeiT. In particular, AutoFormer-tiny/small/base achieve 74.7%/81.7%/82.4% top-1 accuracy on ImageNet with 5.7M/22.9M/53.7M parameters, respectively. Lastly, we verify the transferability of AutoFormer by providing the performance on downstream benchmarks and distillation experiments. Code and models are available at https://github.com/microsoft/Cream.

Visual tracking has achieved considerable progress in recent years. However, current research in the field mainly focuses on tracking of opaque objects, while little attention is paid to transparent object tracking. In this paper, we make the first attempt in exploring this problem by proposing a Transparent Object Tracking Benchmark (TOTB). Specifically, TOTB consists of 225 videos (86K frames) from 15 diverse transparent object categories. Each sequence is manually labeled with axis-aligned bounding boxes. To the best of our knowledge, TOTB is the first benchmark dedicated to transparent object tracking. In order to understand how existing trackers perform and to provide comparison for future research on TOTB, we extensively evaluate 25 state-of-the-art tracking algorithms. The evaluation results exhibit that more efforts are needed to improve transparent object tracking. Besides, we observe some nontrivial findings from the evaluation that are discrepant with some common beliefs in opaque object tracking. For example, we find that deeper features are not always good for improvements. Moreover, to encourage future research, we introduce a novel tracker, named TransATOM, which leverages transparency features for tracking and surpasses all 25 evaluated approaches by a large margin. By releasing TOTB, we expect to facilitate future research and application of transparent object tracking in both the academia and industry. The TOTB and evaluation results as well as TransATOM are available at https: //hengfan2010.github.io/projects/TOTB/.

Accurate camera pose estimation or global camera re-localization is a core component in Structure-from-Motion (SfM) and SLAM systems. Given pair-wise relative camera poses, pose-graph optimization (PGO) involves solving for an optimized set of globally-consistent absolute camera poses. In this work, we propose a novel PGO scheme fueled by graph neural networks (GNN), namely PoGO-Net, to conduct the absolute camera pose regression leveraging multiple rotation averaging (MRA). Specifically, PoGO-Net takes a noisy view-graph as the input, where the nodes and edges are designed to encode the geometric constraints and local graph consistency. Besides, we address the outlier edge removal by exploiting an implicit edge-dropping scheme where the noisy or corrupted edges are effectively filtered out with parameterized networks. Furthermore, we introduce a joint loss function embedding MRA formulation such that the robust inference is capable of achieving real-time performances even for large-scale scenes. Our proposed network is trained end-to-end on public benchmarks, outperforming state-of-the-art approaches in extensive experiments that demonstrate the efficiency and robustness of our proposed network.

Hypergraph matching is a useful tool to find feature correspondence by considering higher-order structural information. Recently, the employment of deep learning has made great progress in the matching of graphs, suggesting its potential for hypergraphs. Hence, in this paper, we present the first, to our best knowledge, unified hypergraph neural network (HNN) solution for hypergraph matching. Specifically, given two hypergraphs to be matched, we first construct an association hypergraph over them and convert the hypergraph matching problem into a node classification problem on the association hypergraph. Then, we design a novel hypergraph neural network to effectively solve the node classification problem. Being end-to-end trainable, our proposed method, named HNN-HM, jointly learns all its components with improved optimization. For evaluation, HNN-HM is tested on various benchmarks and shows a clear advantage over state-of-the-arts.

As a fundamental building block in computer vision, edges can be categorised into four types according to the discontinuity in surface-Reflectance, Illumination, surface-Normal or Depth. While great progress has been made in detecting generic or individual types of edges, it remains under-explored to comprehensively study all four edge types together. In this paper, we propose a novel neural network solution, RINDNet, to jointly detect all four types of edges. Taking into consideration the distinct attributes of each type of edges and the relationship between them, RINDNet learns effective representations for each of them and works in three stages. In stage I, RINDNet uses a common backbone to extract features shared by all edges. Then in stage II it branches to prepare discriminative features for each edge type by the corresponding decoder. In stage III, an independent decision head for each type aggregates the features from previous stages to predict the initial results. Additionally, an attention module learns attention maps for all types to capture the underlying relations between them, and these maps are combined with initial results to generate the final edge detection results. For training and evaluation, we construct the first public benchmark, BSDS-RIND, with all four types of edges carefully annotated. In our experiments, RINDNet yields promising results in comparison with state-of-the-art methods. Additional analysis is presented in supplementary material.

While deep neural networks have shown impressive performance in many tasks, they are fragile to carefully de-signed adversarial attacks. We propose a novel adversarial training-based model by Attention Guided Knowledge Distillation and Bi-directional Metric Learning (AGKD-BML). The attention knowledge is obtained from a weight-fixed model trained on a clean dataset, referred to as a teacher model, and transferred to a model that is under training on adversarial examples (AEs), referred to as a student model. In this way, the student model is able to focus on the correct region, as well as correcting the intermediate features corrupted by AEs to eventually improve the model accuracy. Moreover, to efficiently regularize the representation in feature space, we propose a bidirectional metric learning. Specifically, given a clean image, it is first attacked to its most confusing class to get the forward AE. A clean image in the most confusing class is then randomly picked and attacked back to the original class to get the backward AE. A triplet loss is then used to shorten the representation distance between original image and its AE, while enlarge that between the forward and backward AEs. We conduct extensive adversarial robustness experiments on two widely used datasets with different attacks. Our proposed AGKD-BML model consistently outperforms the state-of-the-art approaches. The code of AGKD-BML will be available at: https://github.com/hongw579/AGKD-BML.

Graph matching aims at establishing correspondence between node sets of given graphs while keeping the consistency between their edge sets. However, outliers in practical scenarios and equivalent learning of edge representations in deep learning methods are still challenging. To address these issues, we present an Edge Attention-adaptive Graph Matching (EAGM) network and a novel description of edge features. EAGM transforms the matching relation between two graphs into a node and edge classification problem over their assignment graph. To explore the potential of edges, EAGM learns edge attention on the assignment graph to 1) reveal the impact of each edge on graph matching, as well as 2) adjust the learning of edge representations adaptively. To alleviate issues caused by the outliers, we describe an edge by aggregating the semantic information over the space spanned by the edge. Such rich information provides clear distinctions between different edges (e.g., inlier-inlier edges vs. inlier-outlier edges), which further distinguishes outliers in the view of their associated edges. Extensive experiments demonstrate that EAGM achieves promising matching quality compared with state-of-the-arts, on cases both with and without outliers. Our source code along with the experiments is available at https://github.com/bestwei/EAGM.

High quality object proposals are crucial in visual tracking algorithms that utilize region proposal network (RPN). Refinement of these proposals, typically by box regression and classification in parallel, has been popularly adopted to boost tracking performance. However, it still meets problems when dealing with complex and dynamic background. Thus motivated, in this paper we introduce an improved proposal refinement module, Cascaded Regression-Align-Classification (CRAC), which yields new state-of-the-art performances on many benchmarks.First, having observed that the offsets from box regression can serve as guidance for proposal feature refinement, we design CRAC as a cascade of box regression, feature alignment and box classification. The key is to bridge box regression and classification via an alignment step, which leads to more accurate features for proposal classification with improved robustness. To address the variation in object appearance, we introduce an identification-discrimination component for box classification, which leverages offline reliable fine-grained template and online rich background information to distinguish the target from background. Moreover, we present pyramid RoIAlign that benefits CRAC by exploiting both the local and global cues of proposals. During inference, tracking proceeds by ranking all refined proposals and selecting the best one. In experiments on seven benchmarks including OTB-2015, UAV123, NfS, VOT-2018, TrackingNet, GOT-10k and LaSOT, our CRACT exhibits very promising results in comparison with state-of-the-art competitors and runs in real-time at 28 fps.

Vision Transformer has shown great visual representation power in substantial vision tasks such as recognition and detection, and thus been attracting fast-growing efforts on manually designing more effective architectures. In this paper, we propose to use neural architecture search to automate this process, by searching not only the architecture but also the search space. The central idea is to gradually evolve different search dimensions guided by their E-T Error computed using a weight-sharing supernet. Moreover, we provide design guidelines of general vision transformers with extensive analysis according to the space searching process, which could promote the understanding of vision transformer. Remarkably, the searched models, named S3 (short for Searching the Search Space), from the searched space achieve superior performance to recently proposed models, such as Swin, DeiT and ViT, when evaluated on ImageNet. The effectiveness of S3 is also illustrated on object detection, semantic segmentation and visual question answering, demonstrating its generality to downstream vision and vision-language tasks. Code and models will be available at https://github.com/microsoft/Cream.

Generic visual object tracking is difficult due to many challenge factors (e.g., occlusion, blur, etc.). Each of these factors may cause serious problems for a tracker, and when they work together can make things even more complicated. Despite a great amount of efforts devoted to understanding the behavior of trackers, reliable and quantifiable ways for studying the per factor tracking behavior remain barely available. Addressing this issue, in this paper we contribute to the community a tracking diagnosis toolkit, TracKlinic, for diagnosis of challenge factors of tracking algorithms.TracKlinic consists of two novel components focusing on the data and analysis aspects, respectively. For the data component, we carefully prepare a set of 2,390 annotated videos, each involving one and only one major challenge factor. When analyzing an algorithm for a specific challenge factor, such one-factor-per-sequence rule greatly inhibits the disturbance from other factors and consequently leads to more faithful analysis. For the analysis component, given the tracking results on all sequences, it investigates the behavior of the tracker under each individual factor and generates the report automatically. With TracKlinic, a thorough study is conducted on ten state-of-the-art trackers on nine challenge factors (including two compound ones). The results suggest that, heavy shape variation and occlusion are the two most challenging factors faced by most trackers. Besides, out-of-view, though does not happen frequently, is often fatal. By sharing TracKlinic1, we expect to make it much easier for diagnosing tracking algorithms, and to thus facilitate developing better ones.

We introduce MART, Motion-Aware Recurrent neural network (MA-RNN) for Tracking, by modeling robust long-term spatial-temporal representation. In particular, we propose a simple, yet effective context-aware displacement attention (CADA) module to capture target motion in videos. By seamlessly integrating CADA into RNN, the proposed MA-RNN can spatially align and aggregate temporal information guided by motion from frame to frame, leading to more effective representation that benefits a tracker from motion when handling occlusion, deformation, viewpoint change etc. Moreover, to deal with scale change, we present a monotonic bounding box regression (mBBR) approach that iteratively predicts regression offsets for target object under the guidance of intersection-over-union (IoU) score, guaranteeing non-decreasing accuracy. In extensive experiments on five benchmarks, including GOT-10k, LaSOT, TC-128, OTB-15 and VOT-19, our tracker MART consistently achieves state-of-the-art results and runs in real-time.

Hyperparameters are numerical pre-sets whose values are assigned prior to the commencement of a learning process. Selecting appropriate hyperparameters is often critical for achieving satisfactory performance in many vision problems, such as deep learning-based visual object tracking. However, it is often difficult to determine their optimal values, especially if they are specific to each video input. Most hyperparameter optimization algorithms tend to search a generic range and are imposed blindly on all sequences. In this paper, we propose a novel dynamical hyperparameter optimization method that adaptively optimizes hyperparameters for a given sequence using an action-prediction network leveraged on continuous deep Q-learning. Since the observation space for object tracking is significantly more complex than those in traditional control problems, existing continuous deep Q-learning algorithms cannot be directly applied. To overcome this challenge, we introduce an efficient heuristic strategy to handle high dimensional state space, while also accelerating the convergence behavior. The proposed algorithm is applied to improve two representative trackers, a Siamese-based one and a correlation-filter-based one, to evaluate its generalizability. Their superior performances on several popular benchmarks are clearly demonstrated. Our source code is available at https://github.com/shenjianbing/dqltracking.

Hashing has recently sparked a great revolution in cross-modal retrieval because of its low storage cost and high query speed. Recent cross-modal hashing methods often learn unified or equal-length hash codes to represent the multi-modal data and make them intuitively comparable. However, such unified or equal-length hash representations could inherently sacrifice their representation scalability because the data from different modalities may not have one-to-one correspondence and could be encoded more efficiently by different hash codes of unequal lengths. To mitigate these problems, this paper exploits a related and relatively unexplored problem: encode the heterogeneous data with varying hash lengths and generalize the cross-modal retrieval in various challenging scenarios. To this end, a generalized and flexible cross-modal hashing framework, termed Matrix Tri-Factorization Hashing (MTFH), is proposed to work seamlessly in various settings including paired or unpaired multi-modal data, and equal or varying hash length encoding scenarios. More specifically, MTFH exploits an efficient objective function to flexibly learn the modality-specific hash codes with different length settings, while synchronously learning two semantic correlation matrices to semantically correlate the different hash representations for heterogeneous data comparable. As a result, the derived hash codes are more semantically meaningful for various challenging cross-modal retrieval tasks. Extensive experiments evaluated on public benchmark datasets highlight the superiority of MTFH under various retrieval scenarios and show its competitive performance with the state-of-the-arts.

) with dynamic eye-tracking data in the unsupervised video object segmentation (UVOS) setting. For the first time, we quantitatively verified the high consistency of visual attention behavior among human observers, and found strong correlation between human attention and explicit primary object judgments during dynamic, task-driven viewing. Such novel observations provide an in-depth insight of the underlying rationale behind video object pattens. Inspired by these findings, we decouple UVOS into two sub-tasks: UVOS-driven Dynamic Visual Attention Prediction (DVAP) in spatiotemporal domain, and Attention-Guided Object Segmentation (AGOS) in spatial domain. Our UVOS solution enjoys three major advantages: 1) modular training without using expensive video segmentation annotations, instead, using more affordable dynamic fixation data to train the initial video attention module and using existing fixation-segmentation paired static/image data to train the subsequent segmentation module; 2) comprehensive foreground understanding through multi-source learning; and 3) additional interpretability from the biologically-inspired and assessable attention. Experiments on four popular benchmarks show that, even without using expensive video object mask annotations, our model achieves compelling performance compared with state-of-the-arts and enjoys fast processing speed (10 fps on a single GPU). Our collected eye-tracking data and algorithm implementations have been made publicly available at https://github.com/wenguanwang/AGS.

Predicting where people look in static scenes, a.k.a visual saliency, has received significant research interest recently. However, relatively less effort has been spent in understanding and modeling visual attention over dynamic scenes. This work makes three contributions to video saliency research. First, we introduce a new benchmark, called DHF1K (Dynamic Human Fixation 1K), for predicting fixations during dynamic scene free-viewing, which is a long-time need in this field. DHF1K consists of 1K high-quality elaborately-selected video sequences annotated by 17 observers using an eye tracker device. The videos span a wide range of scenes, motions, object types and backgrounds. Second, we propose a novel video saliency model, called ACLNet (Attentive CNN-LSTM Network), that augments the CNN-LSTM architecture with a supervised attention mechanism to enable fast end-to-end saliency learning. The attention mechanism explicitly encodes static saliency information, thus allowing LSTM to focus on learning a more flexible temporal saliency representation across successive frames. Such a design fully leverages existing large-scale static fixation datasets, avoids overfitting, and significantly improves training efficiency and testing performance. Third, we perform an extensive evaluation of the state-of-the-art saliency models on three datasets : DHF1K, Hollywood-2, and UCF sports. An attribute-based analysis of previous saliency models and cross-dataset generalization are also presented. Experimental results over more than 1.2K testing videos containing 400K frames demonstrate that ACLNet outperforms other contenders and has a fast processing speed (40 fps using a single GPU). Our code and all the results are available at https://github.com/wenguanwang/DHF1K.

Existing projector-camera calibration methods typically warp keypoints from a camera image to a projector image using estimated homographies and often suffer from errors in camera parameters and noises due to imperfect planarity of the calibration target. This article proposes a practical and robust projector-camera calibration system that explicitly deals with these challenges. First, a graph-theory-based correspondence algorithm is built on top of a color-coded spatial structured light (SL) pattern. Such SL correspondences are then used for a coarse projector-camera calibration. To gain more robustness against noises from an imperfect planar calibration board, we develop a bundle adjustment algorithm to jointly optimize the estimated projector-camera parameters and the correspondences’ coordinates. Moreover, our system requires only one shot of an SL pattern for each calibration board pose, which is much more practical than multishot solutions. Comprehensive experimental validation is conducted on both synthetic and real data sets, and our method clearly outperforms the existing methods in all experiments. For the benefit of the society, a practical open-source software with graphical user interface (GUI) of the developed system is publicly available athttps://github.com/bingyaohuang/single-shot-pro-cam-calib.Note to Practitioners—The proposed method is motivated by two challenges in industrial structured light (SL) system calibration: 1) robustness against imperfect planarity of the calibration target and 2) the number of SL projections per pose. In many industrial SL-based 3-D reconstruction systems, the calibration accuracy greatly affects the reconstruction reliability. Our SL calibration system explicitly deals with calibration target’s imperfect planarity and thus outperforms the existing methods in terms of system accuracy. Another advantage of our SL calibration system is single-shot-per-pose, allowing fast recalibration and reducing the decoding error due to slight pattern misalignment in multishot methods[37]. In addition, we release the open-source calibration software with a graphical user interface (GUI), with which calibration and sparse 3-D reconstruction can be easily performed without any further instructions. Moreover, considering the complex calibration environment and setup, we make the camera and projector imaging parameters, such as exposure, brightness, and contrast, adjustable through widgets and preview. Finally, a limitation of our color-coded SL system is its sensitivity to environment lighting and target texture. This problem may be solved by projector photometric compensation[16],[18],[19],[39].

Classifying the sub-categories of an object from the same super-category (e.g., bird species and cars) in fine-grained visual classification (FGVC) highly relies on discriminative feature representation and accurate region localization. Existing approaches mainly focus on distilling information from high-level features. In this article, by contrast, we show that by integrating low-level information (e.g., color, edge junctions, texture patterns), performance can be improved with enhanced feature representation and accurately located discriminative regions. Our solution, named Attention Pyramid Convolutional Neural Network (AP-CNN), consists of 1) a dual pathway hierarchy structure with a top-down feature pathway and a bottom-up attention pathway, hence learning both high-level semantic and low-level detailed feature representation, and 2) an ROI-guided refinement strategy with ROI-guided dropblock and ROI-guided zoom-in operation, which refines features with discriminative local regions enhanced and background noises eliminated. The proposed AP-CNN can be trained end-to-end, without the need of any additional bounding box/part annotation. Extensive experiments on three popularly tested FGVC datasets (CUB-200-2011, Stanford Cars, and FGVC-Aircraft) demonstrate that our approach achieves state-of-the-art performance. Models and code are available at https://github.com/PRIS-CV/AP-CNN_Pytorch-master.

Existing RGB-D Salient Object Detection (SOD) methods take advantage of depth cues to improve the detection accuracy, while pay insufficient attention to the quality of depth information. In practice, a depth map is often with uneven quality and sometimes suffers from distractors, due to various factors in the acquisition procedure. In this article, to mitigate distractors in depth maps and highlight salient objects in RGB images, we propose a Hierarchical Alternate Interactions Network (HAINet) for RGB-D SOD. Specifically, HAINet consists of three key stages: feature encoding, cross-modal alternate interaction, and saliency reasoning. The main innovation in HAINet is the Hierarchical Alternate Interaction Module (HAIM), which plays a key role in the second stage for cross-modal feature interaction. HAIM first uses RGB features to filter distractors in depth features, and then the purified depth features are exploited to enhance RGB features in turn. The alternate RGB-depth-RGB interaction proceeds in a hierarchical manner, which progressively integrates local and global contexts within a single feature scale. In addition, we adopt a hybrid loss function to facilitate the training of HAINet. Extensive experiments on seven datasets demonstrate that our HAINet not only achieves competitive performance as compared with 19 relevant state-of-the-art methods, but also reaches a real-time processing speed of 43 fps on a single NVIDIA Titan X GPU. The code and results of our method are available at https://github.com/MathLee/HAINet.

As a natural way for human-computer interaction, fixation provides a promising solution for interactive image segmentation. In this paper, we focus on Personal Fixations-based Object Segmentation (PFOS) to address issues in previous studies, such as the lack of appropriate dataset and the ambiguity in fixations-based interaction. In particular, we first construct a new PFOS dataset by carefully collecting pixel-level binary annotation data over an existing fixation prediction dataset, such dataset is expected to greatly facilitate the study along the line. Then, considering characteristics of personal fixations, we propose a novel network based on Object Localization and Boundary Preservation (OLBP) to segment the gazed objects. Specifically, the OLBP network utilizes an Object Localization Module (OLM) to analyze personal fixations and locates the gazed objects based on the interpretation. Then, a Boundary Preservation Module (BPM) is designed to introduce additional boundary information to guard the completeness of the gazed objects. Moreover, OLBP is organized in the mixed bottom-up and top-down manner with multiple types of deep supervision. Extensive experiments on the constructed PFOS dataset show the superiority of the proposed OLBP network over 17 state-of-the-art methods, and demonstrate the effectiveness of the proposed OLM and BPM components. The constructed PFOS dataset and the proposed OLBP network are available at https://github.com/MathLee/OLBPNet4PFOS.

In recent years, multi-view learning has emerged as a promising approach for 3D shape recognition, which identifies a 3D shape based on its 2D views taken from different viewpoints. Usually, the correspondences inside a view or across different views encode the spatial arrangement of object parts and the symmetry of the object, which provide useful geometric cues for recognition. However, such view correspondences have not been explicitly and fully exploited in existing work. In this paper, we propose a correspondence-aware representation (CAR) module, which explicitly finds potential intra-view correspondences and cross-view correspondences via k NN search in semantic space and then aggregates the shape features from the correspondences via learned transforms. Particularly, the spatial relations of correspondences in terms of their viewpoint positions and intra-view locations are taken into account for learning correspondence-aware features. Incorporating the CAR module into a ResNet-18 backbone, we propose an effective deep model called CAR-Net for 3D shape classification and retrieval. Extensive experiments have demonstrated the effectiveness of the CAR module as well as the excellent performance of the CAR-Net.

Spatial misalignment caused by variations in poses and viewpoints is one of the most critical issues that hinder the performance improvement in existing person re-identification (Re-ID) algorithms. Although it is straightforward to explore correspondence learning algorithms for alignment, online learning is intractable for negative pairs due to the intrinsic visual difference between negative pairs and efficiency concern. To address this problem, in this paper, we present a robust and efficient graph correspondence transfer (REGCT) approach for explicit spatial alignment in Re-ID. Specifically, we propose the off-line correspondence learning and on-line correspondence transfer framework. During training, patch-wise correspondences between positive training pairs are established via graph matching. By exploiting both spatial and visual contexts of human appearance in graph matching, meaningful semantic correspondences can be obtained. During testing, the off-line learned patch-wise correspondence templates are transferred to test pairs with similar pose-pair configurations for local feature distance calculation. To enhance the robustness of correspondence transfer, we design a novel pose context descriptor to accurately model human body configurations, and present an approach to measure the similarity between a pair of pose context descriptors. Meanwhile, to improve testing efficiency, we propose a correspondence template ensemble method using the voting mechanism, which significantly reduces the amount of patch-wise matchings involved in distance calculation. With the aforementioned strategies, the REGCT model can effectively and efficiently handle the spatial misalignment problem in Re-ID. Extensive experiments on five challenging benchmarks, including VIPeR, Road, PRID450S, 3DPES, and CUHK01, evidence the superior performance of REGCT over other state-of-the-art approaches.

Deep networks have been used for semantic segmentation tasks on scenes of outdoor environments with increasing popularity. However, the majority of existing work centers on daytime scenes with favorable illumination and weather conditions, and relies on supervision with pixel-level annotations. This paper seeks to address the problem of semantic segmentation for rainy, night-time scenes without using pixel-level annotations. We introduce a near scene semantic approach that uses images of daytime scenes as a bridge for transferring knowledge from pre-trained segmentation models to rainy night images. Specifically, we first present near scene oriented Representation Adaptation (RA) to reduce the domain shift on the representation level. Next, we adapt the segmentation model from the daytime scenario, under varying weather conditions, to the rainy night scenario by using near scene oriented Segmentation Space Adaptation (SSA). Consequently, this further reduces the impact of the domain shift on the segmentation space level. For evaluation, we created a new dataset containing 7000 distinct daytime-night-time image pairs of near scenes obtained by a webcam, and 5266 daytime-rainy night image pairs collected by a car-mounted camera. In addition, we carefully annotated 226 rainy night images with classes defined in Cityscapes. The experimental results clearly demonstrate the advantage of the proposed algorithm.

Image-based relighting, projector compensation and depth/normal reconstruction are three important tasks of projector-camera systems (ProCams) and spatial augmented reality (SAR). Although they share a similar pipeline of finding projector-camera image mappings, in tradition, they are addressed independently, sometimes with different prerequisites, devices and sampling images. In practice, this may be cumbersome for SAR applications to address them one-by-one. In this paper, we propose a novel end-to-end trainable model named DeProCams to explicitly learn the photometric and geometric mappings of ProCams, and once trained, DeProCams can be applied simultaneously to the three tasks. DeProCams explicitly decomposes the projector-camera image mappings into three subprocesses: shading attributes estimation, rough direct light estimation and photorealistic neural rendering. A particular challenge addressed by DeProCams is occlusion, for which we exploit epipolar constraint and propose a novel differentiable projector direct light mask. Thus, it can be learned end-to-end along with the other modules. Afterwards, to improve convergence, we apply photometric and geometric constraints such that the intermediate results are plausible. In our experiments, DeProCams shows clear advantages over previous arts with promising quality and meanwhile being fully differentiable. Moreover, by solving the three tasks in a unified model, DeProCams waives the need for additional optical devices, radiometric calibrations and structured light.

In existing CNN based detectors, the backbone network is a very important component for basic feature1 extraction, and the performance of the detectors highly depends on it. In this paper, we aim to achieve better detection performance by building a more powerful backbone from existing ones like ResNet and ResNeXt. Specifically, we propose a novel strategy for assembling multiple identical backbones by composite connections between the adjacent backbones, to form a more powerful backbone named Composite Backbone Network (CBNet). In this way, CBNet iteratively feeds the output features of the previous backbone, namely high-level features, as part of input features to the succeeding backbone, in a stage-by-stage fashion, and finally the feature maps of the last backbone (named Lead Backbone) are used for object detection. We show that CBNet can be very easily integrated into most state-of-the-art detectors and significantly improve their performances. For example, it boosts the mAP of FPN, Mask R-CNN and Cascade R-CNN on the COCO dataset by about 1.5 to 3.0 points. Moreover, experimental results show that the instance segmentation results can be improved as well. Specifically, by simply integrating the proposed CBNet into the baseline detector Cascade Mask R-CNN, we achieve a new state-of-the-art result on COCO dataset (mAP of 53.3) with a single model, which demonstrates great effectiveness of the proposed CBNet architecture. Code will be made available at https://github.com/PKUbahuangliuhe/CBNet.

Learning-based approaches to graph matching have been developed and explored for more than a decade, have grown rapidly in scope and popularity in recent years. However, previous learning-based algorithms, with or without deep learning strategy, mainly focus on the learning of node and/or edge affinities generation, and pay less attention on the learning of the combinatorial solver. In this paper we propose a fully trainable framework for graph matching, in which learning of affinities and solving for combinatorial optimization are not explicitly separated as in many previous arts. We firstly convert the problem of building node correspondences between two input graphs to the problem of selecting reliable nodes from a constructed assignment graph. Subsequently, the graph network block module is adopted to perform computation on the graph to form structured representations for each node. It finally predicts a label for each node that is used for node classification, and the training is performed under the supervision of both permutation differences and the one-to-one matching constraints. The proposed method is evaluated on four public benchmarks in comparison with several state-of-the-art algorithms, and the experimental results illustrate its excellent performance.

Rapid progress has been witnessed for human-object interaction (HOI) recognition, but most existing models are confined to single-stage reasoning pipelines. Considering the intrinsic complexity of the task, we introduce a cascade architecture for a multi-stage, coarse-to-fine HOI understanding. At each stage, an instance localization network progressively refines HOI proposals and feeds them into an interaction recognition network. Each of the two networks is also connected to its predecessor at the previous stage, enabling cross-stage information propagation. The interaction recognition network has two crucial parts: a relation ranking module for high-quality HOI proposal selection and a triple-stream classifier for relation prediction. With our carefully-designed human-centric relation features, these two modules work collaboratively towards effective interaction understanding. Further beyond relation detection on a bounding-box level, we make ourframework flexible to perform fine-grained pixel-wise relation segmentation; this provides a new glimpse into better relation modeling. Our approach reached the 1st place in the ICCV2019 Person in Context Challenge, on both relation detection and segmentation tasks. It also shows promising results on V-COCO.

Existing CNN-based methods for semantic segmentation heavily depend on multi-scale features to meet the requirements of both semantic comprehension and detail preservation. State-of-the-art segmentation networks widely exploit conventional scale-transfer operations, i.e., up-sampling and down-sampling to learn multi-scale features. In this work, we find that these operations lead to scale-confused features and suboptimal performance because they are spatial-invariant and directly transit all feature information cross scales without spatial selection. To address this issue, we propose the Gated Scale-Transfer Operation (GSTO) to properly transit spatial-filtered features to another scale. Specifically, GSTO can work either with or without extra supervision. Unsupervised GSTO is learned from the feature itself while the supervised one is guided by the supervised probability matrix. Both forms of GSTO are lightweight and plug-and-play, which can be flexibly integrated into networks or modules for learning better multi-scale features. In particular, by plugging GSTO into HRNet, we get a more powerful backbone (namely GSTO-HRNet) for pixel labeling, and it achieves new state-of-the-art results on multiple benchmarks for semantic segmentation including Cityscapes, LIP, and Pascal Context, with a negligible extra computational cost. Moreover, experiment results demonstrate that GSTO can also significantly boost the performance of multi-scale feature aggregation modules like PPM and ASPP.

In this paper, we present ARCHIE, a framework for testing augmented reality applications in the wild. ARCHIE collects user feedback and system state data in situ to help developers identify and debug issues important to testers. It also supports testing of multiple application versions (called "profiles") in a single evaluation session, prioritizing those versions which the tester finds more appealing. To evaluate ARCHIE, we implemented four distinct test case applications and used these applications to examine the performance overhead and context switching cost of incorporating our framework into a pre-existing code base. With these, we demonstrate that ARCHIE provides no significant overhead for AR applications, and introduces at most 2% processing overhead when switching among large groups of testable profiles.

Abstract The multi-dimensional assignment problem is universal for data association analysis such as data association-based visual multi-object tracking and multi-graph matching. In this paper, multi-dimensional assignment is formulated as a rank-1 tensor approximation problem. A dualL1-normalized context/hyper-context aware tensor power iteration optimization method is proposed. The method is applied to multi-object tracking and multi-graph matching. In the optimization method, tensor power iteration with the dual unit norm enables the capture of information across multiple sample sets. Interactions between sample associations are modeled as contexts or hyper-contexts which are combined with the global affinity into a unified optimization. The optimization is flexible for accommodating various types of contextual models. In multi-object tracking, the global affinity is defined according to the appearance similarity between objects detected in different frames. Interactions between objects are modeled as motion contexts which are encoded into the global association optimization. The tracking method integrates high order motion information and high order appearance variation. The multi-graph matching method carries out matching over graph vertices and structure matching over graph edges simultaneously. The matching consistency across multi-graphs is based on the high-order tensor optimization. Various types of vertex affinities and edge/hyper-edge affinities are flexibly integrated. Experiments on several public datasets, such as the MOT16 challenge benchmark, validate the effectiveness of the proposed methods.

This paper presents a new Gaussian Processes (GPs)-based particle filter tracking framework. The framework non-trivially extends Gaussian process regression (GPR) to transfer learning, and, following the tracking-by-fusion strategy, integrates closely two tracking components, namely a GPs component and a CFs one. First, the GPs component analyzes and models the probability distribution of the object appearance by exploiting GPs. It categorizes the labeled samples into auxiliary and target ones, and explores unlabeled samples in transfer learning. The GPs component thus captures rich appearance information over object samples across time. On the other hand, to sample an initial particle set in regions of high likelihood through the direct simulation method in particle filtering, the powerful yet efficient correlation filters (CFs) are integrated, leading to the CFs component. In fact, the CFs component not only boosts the sampling quality, but also benefits from the GPs component, which provides re-weighted knowledge as latent variables for determining the impact of each correlation filter template from the auxiliary samples. In this way, the transfer learning based fusion enables effective interactions between the two components. Superior performance on four object tracking benchmarks (OTB-2015, Temple-Color, and VOT2015/2016), and in comparison with baselines and recent state-of-the-art trackers, has demonstrated clearly the effectiveness of the proposed framework.

Filtering images is required by numerous multimedia, computer vision and graphics tasks. Despite diverse goals of different tasks, making effective rules is key to the filtering performance. Linear translation-invariant filters with manually designed kernels have been widely used. However, their performance suffers from content-blindness. To mitigate the content-blindness, a family of filters, called joint/guided filters, have attracted a great amount of attention from the community. The main drawback of most joint/guided filters comes from the ignorance of structural inconsistency between the reference and target signals like color, infrared, and depth images captured under different conditions. Simply adopting such guidelines very likely leads to unsatisfactory results. To address the above issues, this paper designs a simple yet effective filter, named mutually guided image filter (muGIF), which jointly preserves mutual structures, avoids misleading from inconsistent structures and smooths flat regions. The proposed muGIF is very flexible, which can work in various modes including dynamic only (self-guided), static/dynamic (reference-guided) and dynamic/dynamic (mutually guided) modes. Although the objective of muGIF is in nature non-convex, by subtly decomposing the objective, we can solve it effectively and efficiently. The advantages of muGIF in effectiveness and flexibility are demonstrated over other state-of-the-art alternatives on a variety of applications. Our code is publicly available at https://sites.google.com/view/xjguo/mugif.

Visual Augmented Reality (AR) typically overlays virtual computer graphics or other virtual contents on the real world videos, attracting much interest from both academic and industrial communities. Although AR techniques on planes are well studied, cylindrical objects are seldom used for augmented reality. In this paper, we propose a new method for 3D reconstruction and 6D pose computation for augmented reality on a cylindrical object. The 6D pose is the relative pose between the camera and the cylindrical object, which is very convenient to make augmented reality. First, we capture some images with a cylindrical object and then reconstruct its 3D model with textures offline by using projective invariance and image contours. Second, according to the 3D model, we track the 6D relative pose between the camera and the cylindrical object online, where we propose a linear P3P RANSAC to remove outliers. Finally, the virtual images are exactly aligned with the cylindrical object in the real world. Experimental results show that the proposed method outperforms the state of the arts in terms of 3D mapping and 6D pose computation on cylindrical objects.

RGB-D based salient object detection (SOD) methods leverage the depth map as a valuable complementary information for better SOD performance. Previous methods mainly resort to exploit the correlation between RGB image and depth map in three fusion domains: input images, extracted features, and output results. However, these fusion strategies cannot fully capture the complex correlation between the RGB image and depth map. Besides, these methods do not fully explore the cross-modal complementarity and the cross-level continuity of information, and treat information from different sources without discrimination. In this paper, to address these problems, we propose a novel Information Conversion Network (ICNet) for RGB-D based SOD by employing the siamese structure with encoder-decoder architecture. To fuse high-level RGB and depth features in an interactive and adaptive way, we propose a novel Information Conversion Module (ICM), which contains concatenation operations and correlation layers. Furthermore, we design a Cross-modal Depth-weighted Combination (CDC) block to discriminate the cross-modal features from different sources and to enhance RGB features with depth features at each level. Extensive experiments on five commonly tested datasets demonstrate the superiority of our ICNet over 15 state-of-theart RGB-D based SOD methods, and validate the effectiveness of the proposed ICM and CDC block.

Recent progresses in visual tracking have greatly improved the tracking performance. However, challenges such as occlusion and view change remain obstacles in real world deployment. A natural solution to these challenges is to use multiple cameras with multiview inputs, though existing systems are mostly limited to specific targets (e.g. human), static cameras, and/or require camera calibration. To break through these limitations, we propose a generic multiview tracking (GMT) framework that allows camera movement, while requiring neither specific object model nor camera calibration. A key innovation in our framework is a cross-camera trajectory prediction network (TPN), which implicitly and dynamically encodes camera geometric relations, and hence addresses missing target issues such as occlusion. Moreover, during tracking, we assemble information across different cameras to dynamically update a novel collaborative correlation filter (CCF), which is shared among cameras to achieve robustness against view change. The two components are integrated into a correlation filter tracking framework, where features are trained offline using existing single view tracking datasets. For evaluation, we first contribute a new generic multiview tracking dataset (GMTD) with careful annotations, and then run experiments on the GMTD and CAMPUS datasets. The proposed GMT algorithm shows clear advantages in terms of robustness over state-of-the-art ones.

The goal of our work is to discover dominant objects in a very general setting where only a single unlabeled image is given. This is far more challenge than typical colocalization or weakly-supervised localization tasks. To tackle this problem, we propose a simple but effective pattern mining-based method, called Object Location Mining (OLM), which exploits the advantages of data mining and feature representation of pretrained convolutional neural networks (CNNs). Specifically, we first convert the feature maps from a pre-trained CNN model into a set of transactions, and then discovers frequent patterns from transaction database through pattern mining techniques. We observe that those discovered patterns, i.e., co-occurrence highlighted regions, typically hold appearance and spatial consistency. Motivated by this observation, we can easily discover and localize possible objects by merging relevant meaningful patterns. Extensive experiments on a variety of benchmarks demonstrate that OLM achieves competitive localization performance compared with the state-of-the-art methods. We also evaluate our approach compared with unsupervised saliency detection methods and achieves competitive results on seven benchmark datasets. Moreover, we conduct experiments on finegrained classification to show that our proposed method can locate the entire object and parts accurately, which can benefit to improving the classification results significantly.

Image alignment/registration/correspondence is a critical prerequisite for many vision-based tasks, and it has been widely studied in computer vision. However, aligning images from different domains, such as cross-weather/season road scenes, remains a challenging problem. Inspired by the success of classic intensity-constancy-based image alignment methods and the modern generative adversarial network (GAN) technology, we propose a cross-weather road scene alignment method called latent generative model with intensity constancy. From a novel perspective, the alignment problem is formulated as a constrained 2D flow optimization problem with latent encoding, which can be decoded into an intensity-constancy image on the latent image manifold. The manifold is parameterized by a pre-trained GAN, which is able to capture statistic characteristics from large datasets. Moreover, we employ the learned manifold to constrain the warped latent image identical to the target image, thereby producing a realistic warping effect. Experimental results on several cross-weather/season road scene datasets demonstrate that our approach can significantly outperform the state-of-the-art methods.

Pavement crack detection is a critical task for insuring road safety. Manual crack detection is extremely time-consuming. Therefore, an automatic road crack detection method is required to boost this progress. However, it remains a challenging task due to the intensity inhomogeneity of cracks and complexity of the background, e.g., the low contrast with surrounding pavements and possible shadows with a similar intensity. Inspired by recent advances of deep learning in computer vision, we propose a novel network architecture, named feature pyramid and hierarchical boosting network (FPHBN), for pavement crack detection. The proposed network integrates context information to low-level features for crack detection in a feature pyramid way, and it balances the contributions of both easy and hard samples to loss by nested sample reweighting in a hierarchical way during training. In addition, we propose a novel measurement for crack detection named average intersection over union (AIU). To demonstrate the superiority and generalizability of the proposed method, we evaluate it on five crack datasets and compare it with the state-of-the-art crack detection, edge detection, and semantic segmentation methods. The extensive experiments show that the proposed method outperforms these methods in terms of accuracy and generalizability. Code and data can be found in https://github.com/fyangneil/pavement-crack-detection.

We propose an automatic framework to recover the illumination of indoor scenes based on a single RGB-D image. Unlike previous works, our method can recover spatially varying illumination without using any lighting capturing devices or HDR information. The recovered illumination can produce realistic rendering results. To model the geometry of the visible and invisible parts of scenes corresponding to the input RGB-D image, we assume that all objects shown in the image are located in a box with six faces and build a planar-based geometry model based on the input depth map. We then present a confidence-scoring based strategy to separate the light sources from the highlight areas. The positions of light sources both in and out of the camera's view are calculated based on the classification result and the recovered geometry model. Finally, an iterative procedure is proposed to calculate the colors of light sources and the materials in the scene. In addition, a data-driven method is used to set constraints on the light source intensities. Using the estimated light sources and geometry model, environment maps at different points in the scene are generated that can model the spatial variance of illumination. The experimental results demonstrate the validity and flexibility of our approach.

Feature pyramids are widely exploited by both the state-of-the-art one-stage object detectors (e.g., DSSD, RetinaNet, RefineDet) and the two-stage object detectors (e.g., Mask RCNN, DetNet) to alleviate the problem arising from scale variation across object instances. Although these object detectors with feature pyramids achieve encouraging results, they have some limitations due to that they only simply construct the feature pyramid according to the inherent multiscale, pyramidal architecture of the backbones which are originally designed for object classification task. Newly, in this work, we present Multi-Level Feature Pyramid Network (MLFPN) to construct more effective feature pyramids for detecting objects of different scales. First, we fuse multi-level features (i.e. multiple layers) extracted by backbone as the base feature. Second, we feed the base feature into a block of alternating joint Thinned U-shape Modules and Feature Fusion Modules and exploit the decoder layers of each Ushape module as the features for detecting objects. Finally, we gather up the decoder layers with equivalent scales (sizes) to construct a feature pyramid for object detection, in which every feature map consists of the layers (features) from multiple levels. To evaluate the effectiveness of the proposed MLFPN, we design and train a powerful end-to-end one-stage object detector we call M2Det by integrating it into the architecture of SSD, and achieve better detection performance than state-of-the-art one-stage detectors. Specifically, on MSCOCO benchmark, M2Det achieves AP of 41.0 at speed of 11.8 FPS with single-scale inference strategy and AP of 44.2 with multi-scale inference strategy, which are the new stateof-the-art results among one-stage detectors. The code will be made available on https://github.com/qijiezhao/M2Det.

Recently, the region proposal networks (RPN) have been combined with the Siamese network for tracking, and shown excellent accuracy with high efficiency. Nevertheless, previously proposed one-stage Siamese-RPN trackers degenerate in presence of similar distractors and large scale variation. Addressing these issues, we propose a multi-stage tracking framework, Siamese Cascaded RPN (C-RPN), which consists of a sequence of RPNs cascaded from deep high-level to shallow low-level layers in a Siamese network. Compared to previous solutions, C-RPN has several advantages: (1) Each RPN is trained using the outputs of RPN in the previous stage. Such process stimulates hard negative sampling, resulting in more balanced training samples. Consequently, the RPNs are sequentially more discriminative in distinguishing difficult background (i.e.,, similar distractors). (2) Multi-level features are fully leveraged through a novel feature transfer block (FTB) for each RPN, further improving the discriminability of C-RPN using both high-level semantic and low-level spatial information. (3) With multiple steps of regressions, C-RPN progressively refines the location and shape of the target in each RPN with adjusted anchor boxes in the previous stage, which makes localization more accurate. C-RPN is trained end-to-end with the multi-task loss function. In inference, C-RPN is deployed as it is, without any temporal adaption, for real-time tracking. In extensive experiments on OTB-2013, OTB-2015, VOT-2016, VOT-2017, LaSOT and TrackingNet, C-RPN consistently achieves state-of-the-art results and runs in real-time.

In this paper, we present LaSOT, a high-quality benchmark for Large-scale Single Object Tracking. LaSOT consists of 1,400 sequences with more than 3.5M frames in total. Each frame in these sequences is carefully and manually annotated with a bounding box, making LaSOT the largest, to the best of our knowledge, densely annotated tracking benchmark. The average video length of LaSOT is more than 2,500 frames, and each sequence comprises various challenges deriving from the wild where target objects may disappear and re-appear again in the view. By releasing LaSOT, we expect to provide the community with a large-scale dedicated benchmark with high quality for both the training of deep trackers and the veritable evaluation of tracking algorithms. Moreover, considering the close connections of visual appearance and natural language, we enrich LaSOT by providing additional language specification, aiming at encouraging the exploration of natural linguistic feature for tracking. A thorough experimental evaluation of 35 tracking algorithms on LaSOT is presented with detailed analysis, and the results demonstrate that there is still a big room for improvements.

Projector photometric compensation aims to modify a projector input image such that it can compensate for disturbance from the appearance of projection surface. In this paper, for the first time, we formulate the compensation problem as an end-to-end learning problem and propose a convolutional neural network, named CompenNet, to implicitly learn the complex compensation function. CompenNet consists of a UNet-like backbone network and an autoencoder subnet. Such architecture encourages rich multi-level interactions between the camera-captured projection surface image and the input image, and thus captures both photometric and environment information of the projection surface. In addition, the visual details and interaction information are carried to deeper layers along the multi-level skip convolution layers. The architecture is of particular importance for the projector compensation task, for which only a small training dataset is allowed in practice. Another contribution we make is a novel evaluation benchmark, which is independent of system setup and thus quantitatively verifiable. Such benchmark is not previously available, to our best knowledge, due to the fact that conventional evaluation requests the hardware system to actually project the final results. Our key idea, motivated from our end-to-end problem formulation, is to use a reasonable surrogate to avoid such projection process so as to be setup-independent. Our method is evaluated carefully on the benchmark, and the results show that our end-to-end learning solution outperforms state-of-the-arts both qualitatively and quantitatively by a significant margin.

This paper conducts a systematic study on the role of visual attention in Unsupervised Video Object Segmentation (UVOS) tasks. By elaborately annotating three popular video segmentation datasets (DAVIS, Youtube-Objects and SegTrack V2) with dynamic eye-tracking data in the UVOS setting, for the first time, we quantitatively verified the high consistency of visual attention behavior among human observers, and found strong correlation between human attention and explicit primary object judgements during dynamic, task-driven viewing. Such novel observations provide an in-depth insight into the underlying rationale behind UVOS. Inspired by these findings, we decouple UVOS into two sub-tasks: UVOS-driven Dynamic Visual Attention Prediction (DVAP) in spatiotemporal domain, and Attention-Guided Object Segmentation (AGOS) in spatial domain. Our UVOS solution enjoys three major merits: 1) modular training without using expensive video segmentation annotations, instead, using more affordable dynamic fixation data to train the initial video attention module and using existing fixation-segmentation paired static/image data to train the subsequent segmentation module; 2) comprehensive foreground understanding through multi-source learning; and 3) additional interpretability from the biologically-inspired and assessable attention. Experiments on popular benchmarks show that, even without using expensive video object mask annotations, our model achieves compelling performance in comparison with state-of-the-arts.

This paper addresses the problem of deformable surface tracking from monocular images. Specifically, we propose a graph-based approach that effectively explores the structure information of the surface to enhance tracking performance. Our approach solves simultaneously for feature correspondence, outlier rejection and shape reconstruction by optimizing a single objective function, which is defined by means of pairwise projection errors between graph structures instead of unary projection errors between matched points. Furthermore, an efficient matching algorithm is developed based on soft matching relaxation. For evaluation, our approach is extensively compared to state-of-the-art algorithms on a standard dataset of occluded surfaces, as well as a newly compiled dataset of different surfaces with rich, weak or repetitive texture. Experimental results reveal that our approach achieves robust tracking results for surfaces with different types of texture, and outperforms other algorithms in both accuracy and efficiency.

Data association-based multiple object tracking (MOT) involves multiple separated modules processed or optimized differently, which results in complex method design and requires non-trivial tuning of parameters. In this paper, we present an end-to-end model, named FAMNet, where Feature extraction, Affinity estimation and Multi-dimensional assignment are refined in a single network. All layers in FAMNet are designed differentiable thus can be optimized jointly to learn the discriminative features and higher-order affinity model for robust MOT, which is supervised by the loss directly from the assignment ground truth. In addition, we integrate single object tracking technique and a dedicated target management scheme into the FAMNet-based tracking system to further recover false negatives and inhibit noisy target candidates generated by the external detector. The proposed method is evaluated on a diverse set of benchmarks including MOT2015, MOT2017, KITTI-Car and UA-DETRAC, and achieves promising performance on all of them in comparison with state-of-the-arts.

Full projector compensation aims to modify a projector input image such that it can compensate for both geometric and photometric disturbance of the projection surface. Traditional methods usually solve the two parts separately, although they are known to correlate with each other. In this paper, we propose the first end-to-end solution, named CompenNet++, to solve the two problems jointly. Our work non-trivially extends CompenNet, which was recently proposed for photometric compensation with promising performance. First, we propose a novel geometric correction subnet, which is designed with a cascaded coarse-to-fine structure to learn the sampling grid directly from photometric sampling images. Second, by concatenating the geometric correction subset with CompenNet, CompenNet++ accomplishes full projector compensation and is end-to-end trainable. Third, after training, we significantly simplify both geometric and photometric compensation parts, and hence largely improves the running time efficiency. Moreover, we construct the first setup-independent full compensation benchmark to facilitate the study on this topic. In our thorough experiments, our method shows clear advantages over previous arts with promising compensation quality and meanwhile being practically convenient.

This paper proposes a human-aware deblurring model that disentangles the motion blur between foreground (FG) humans and background (BG). The proposed model is based on a triple-branch encoder-decoder architecture. The first two branches are learned for sharpening FG humans and BG details, respectively; while the third one produces global, harmonious results by comprehensively fusing multi-scale deblurring information from the two domains. The proposed model is further endowed with a supervised, human-aware attention mechanism in an end-to-end fashion. It learns a soft mask that encodes FG human information and explicitly drives the FG/BG decoder-branches to focus on their specific domains. Above designs lead to a fully differentiable motion deblurring network, which can be trained end-to-end. To further benefit the research towards Human-aware Image Deblurring, we introduce a large-scale dataset, named HIDE, which consists of 8,422 blurry and sharp image pairs with 65,784 densely annotated FG human bounding boxes. HIDE is specifically built to span a broad range of scenes, human object sizes, motion patterns, and background complexities. Extensive experiments on public benchmarks and our dataset demonstrate that our model performs favorably against the state-of-the-art motion deblurring methods, especially in capturing semantic details.

Detecting objects in aerial images is challenging for at least two reasons: (1) target objects like pedestrians are very small in pixels, making them hardly distinguished from surrounding background; and (2) targets are in general sparsely and non-uniformly distributed, making the detection very inefficient. In this paper, we address both issues inspired by observing that these targets are often clustered. In particular, we propose a Clustered Detection (ClusDet) network that unifies object clustering and detection in an end-to-end framework. The key components in ClusDet include a cluster proposal sub-network (CPNet), a scale estimation sub-network (ScaleNet), and a dedicated detection network (DetecNet). Given an input image, CPNet produces object cluster regions and ScaleNet estimates object scales for these regions. Then, each scale-normalized cluster region is fed into DetecNet for object detection. ClusDet has several advantages over previous solutions: (1) it greatly reduces the number of chips for final object detection and hence achieves high running time efficiency, (2) the cluster-based scale estimation is more accurate than previously used single-object based ones, hence effectively improves the detection for small objects, and (3) the final DetecNet is dedicated for clustered regions and implicitly models the prior context information so as to boost detection accuracy. The proposed method is tested on three popular aerial image datasets including VisDrone, UAVDT and DOTA. In all experiments, ClusDet achieves promising performance in comparison with state-of-the-art detectors.

Recent progresses in model-free single object tracking (SOT) algorithms have largely inspired applying SOT to multi-object tracking (MOT) to improve the robustness as well as relieving dependency on external detector. However, SOT algorithms are generally designed for distinguishing a target from its environment, and hence meet problems when a target is spatially mixed with similar objects as observed frequently in MOT. To address this issue, in this paper we propose an instance-aware tracker to integrate SOT techniques for MOT by encoding awareness both within and between target models. In particular, we construct each target model by fusing information for distinguishing target both from background and other instances (tracking targets). To conserve uniqueness of all target models, our instance-aware tracker considers response maps from all target models and assigns spatial locations exclusively to optimize the overall accuracy. Another contribution we make is a dynamic model refreshing strategy learned by a convolutional neural network. This strategy helps to eliminate initialization noise as well as to adapt to variation of target size and appearance. To show the effectiveness of the proposed approach, it is evaluated on the popular MOT15 and MOT16 challenge benchmarks. On both benchmarks, our approach achieves the best overall performances in comparison with published results.

Recurrent neural networks (RNNs) have shown the ability to improve scene parsing through capturing long-range dependencies among image units. In this paper, we propose dense RNNs for scene labeling by exploring various long-range semantic dependencies among image units. Different from existing RNN based approaches, our dense RNNs are able to capture richer contextual dependencies for each image unit by enabling immediate connections between each pair of image units, which significantly enhances their discriminative power. Besides, to select relevant dependencies and meanwhile to restrain irrelevant ones for each unit from dense connections, we introduce an attention model into dense RNNs. The attention model allows automatically assigning more importance to helpful dependencies while less weight to unconcerned dependencies. Integrating with convolutional neural networks (CNNs), we develop an end-to-end scene labeling system. Extensive experiments on three large-scale benchmarks demonstrate that the proposed approach can improve the baselines by large margins and outperform other state-of-the-art algorithms.

We study the problem of photo cropping, which aims to find a cropping window of an input image to preserve as much as possible its important parts while being aesthetically pleasant. Seeking a deep learning-based solution, we design a neural network that has two branches for attention box prediction (ABP) and aesthetics assessment (AA), respectively. Given the input image, the ABP network predicts an attention bounding box as an initial minimum cropping window, around which a set of cropping candidates are generated with little loss of important information. Then, the AA network is employed to select the final cropping window with the best aesthetic quality among the candidates. The two sub-networks are designed to share the same full-image convolutional feature map, and thus are computationally efficient. By leveraging attention prediction and aesthetics assessment, the cropping model produces high-quality cropping results, even with the limited availability of training data for photo cropping. The experimental results on benchmark datasets clearly validate the effectiveness of the proposed approach. In addition, our approach runs at 5 fps, outperforming most previous solutions. The code and results are available at: https://github.com/shenjianbing/DeepCropping.

Visual object tracking has played a crucial role in computer vision with many applications. Being intensively studied in recent decades, visual tracking has witnessed great advances in either speed (e.g., with correlation filters) or accuracy (e.g., with deep features). Real-time and high accuracy tracking algorithms, nevertheless, remain scarce. In this paper we study the problem from a new perspective and present a novel parallel tracking and verifying (PTAV) framework, by taking advantage of the ubiquity of multi-thread techniques and borrowing ideas from the success of parallel tracking and mapping in visual SLAM. The proposed PTAV framework typically consists of two components, a (base) tracker T and a verifier V, working in parallel on two separate threads. The tracker T aims at providing a super real-time tracking inference and is expected to perform well most of the time; by contrast, the verifier V validates the tracking results and corrects T when needed. The key innovation is that, V does not work on every frame but only upon the requests from T; on the other end, T may adjust the tracking according to the feedback from V. With such collaboration, PTAV enjoys both high efficiency provided by T and strong discriminative power by V. Meanwhile, in order to adapt V to object appearance changes, we maintain a dynamic target template pool for adaptive verification, resulting in further improvement. In our extensive experiments on OTB2015, TC128, UAV20L and VOT2016, PTAV achieves top tracking accuracy among all real-time trackers, and in fact even outperforms many deep learning based algorithms. Moreover, as a general framework, PTAV is very flexible with great potentials for future improvement and generalization.

Differential features are popularly used in computer vision tasks, such as object detection. In this paper, we revisit these features from a functional approximation perspective. In particular, we view an image as a 2-D functional and investigate its Taylor series approximation. Differential features are derived from the approximation coefficients and, therefore, are naturally collected for appearance representation. Thus motivated, we propose to use the zeroth-, first-, and second-order differential features for pedestrian detection and call such features Taylor feature transform (TAFT). In practice, the TAFT features are computed by discrete sampling to address scale issues and meanwhile achieve computational efficiency. In addition, orientation insensitivity is handled by using directional versions of differentials. When applied to pedestrian detection, the TAFT is sampled on grid pixels and calculated from multiple channels following previous solutions. In our extensive experiments on the INRIA, Caltech, TUD-Brussel, and KITTI data sets, the TAFT achieves state-of-the-art results. It outperforms all handcrafted features and performs on par with many deep-learning solutions. Moreover, when a low false-positive rate is requested, the TAFT generates results that are better than or comparable to the state-of-the-art deep learning-based methods. Meanwhile, our implementation runs at 33 fps for 640×480 images without GPU, making TAFT favorable in many practical scenarios.

We design a pedestrian proposal and refining system tailored for fast pedestrian detection. The pedestrian proposal is based on pixel differential feature (PDF), which is a light weighted feature with a high recall rate. For the pedestrian refining, we propose an aggregated region feature (ARF) to distill the co-existing dominant pixel differential patterns in a local region to reject hard false positives. Albeit discriminative, ARF largely relies on the size of anchored regions and the scale of the PDF, which hinders its performance in real-world applications. Although multi-scale PDF with spatial pyramid somewhat alleviates this problem, it is computationally expensive and thus infeasible in practice. To address this issue, we further propose a directional radius pooling method to extract discriminative information in each orientation of PDF while reducing the feature dimensionality with a more compact size. The pedestrian proposal and refining framework is built on the shared pixel differential feature map which is very computationally efficient. More specifically, a set of pedestrian proposals generated from the single-scale PDF are first obtained in images. Second, multi-scale ARF in spatial pyramid is used to fuse information from different scales and spatial resolutions for anchor regions. Third, the directional radius pooling method is proposed to extract dominant information of each orientation in the anchor regions. The pedestrian proposal and refining are finally integrated for accurate pedestrian detection. The extensive experimental evaluations on five public benchmarks show that our method achieves state-of-the-art results while running at 18 fps for $480\times640$ images.

In this paper, we propose a novel visual-inertial odometry (VIO) approach that adopts structural regularity in man-made environments. Instead of using Manhattan world assumption, we use Atlanta world model to describe such regularity. An Atlanta world is a world that contains multiple local Manhattan worlds with different heading directions. Each local Manhattan world is detected on the fly, and their headings are gradually refined by the state estimator when new observations are received. With full exploration of structural lines that aligned with each local Manhattan worlds, our VIO method becomes more accurate and robust, as well as more flexible to different kinds of complex man-made environments. Through benchmark tests and real-world tests, the results show that the proposed approach outperforms existing visual-inertial systems in large-scale man-made environments.

In this paper, we propose a graph correspondence transfer (GCT) approach for person re-identification. Unlike existing methods, the GCT model formulates person re-identification as an off-line graph matching and on-line correspondence transferring problem. In specific, during training, the GCT model aims to learn off-line a set of correspondence templates from positive training pairs with various pose-pair configurations via patch-wise graph matching. During testing, for each pair of test samples, we select a few training pairs with the most similar pose-pair configurations as references, and transfer the correspondences of these references to test pair for feature distance calculation. The matching score is derived by aggregating distances from different references. For each probe image, the gallery image with the highest matching score is the re-identifying result. Compared to existing algorithms, our GCT can handle spatial misalignment caused by large variations in view angles and human poses owing to the benefits of patch-wise graph matching. Extensive experiments on five benchmarks including VIPeR, Road, PRID450S, 3DPES and CUHK01 evidence the superior performance of GCT model over other state-of-the-art methods.

Tracking planar objects has a wide range of applications in robotics. Conventional template tracking algorithms, however, often fail to observe fast object motion or drift significantly after a period of time, due to drastic object appearance change. To address such challenges, we propose a novel constrained confidence matching algorithm for motion estimation and a robust Kalman filter for template updating. Integrated with an accurate occlusion detector, our approach achieves accurate motion estimation in presence of partial occlusion, by excluding occluded pixels from computation of motion parameters. Furthermore, the proposed Kalman filter employs a novel control-input model to handle the object appearance change, which brings our tracker high robustness against sudden illumination change and heavy motion blur. For evaluation, we compare the proposed tracker with several state-of-the-art planar object trackers on two public benchmark datasets. Experimental results show that our algorithm achieves robust tracking results against various environmental variations, and outperforms baseline algorithms remarkably on both datasets.

Planar object tracking is an actively studied problem in vision-based robotic applications. While several benchmarks have been constructed for evaluating state-of-the-art algorithms, there is a lack of video sequences captured in the wild rather than in constrained laboratory environment. In this paper, we present a carefully designed planar object tracking benchmark containing 210 videos of 30 planar objects sampled in the natural environment. In particular, for each object, we shoot seven videos involving various challenging factors, namely scale change, rotation, perspective distortion, motion blur, occlusion, out-of-view, and unconstrained. The ground truth is carefully annotated semi-manually to ensure the quality. Moreover, eleven state-of-the-art algorithms are evaluated on the benchmark using two evaluation metrics, with detailed analysis provided for the evaluation results. We expect the proposed benchmark to benefit future studies on planar object tracking.

Matching-based algorithms have been commonly used in planar object tracking. They often model a planar object as a set of keypoints, and then find correspondences between keypoint sets via descriptor matching. In previous work, unary constraints on appearances or locations are usually used to guide the matching. However, these approaches rarely utilize structure information of the object, and are thus suffering from various perturbation factors. In this paper, we proposed a graph-based tracker, named Gracker, which is able to fully explore the structure information of the object to enhance tracking performance. We model a planar object as a graph, instead of a simple collection of keypoints, to represent its structure. Then, we reformulate tracking as a sequential graph matching process, which establishes keypoint correspondence in a geometric graph matching manner. For evaluation, we compare the proposed Gracker with state-of-the-art planar object trackers on three benchmark datasets: two public ones and a newly collected one. Experimental results show that Gracker achieves robust tracking results against various environmental variations, and outperforms other algorithms in general on the datasets.

Graph matching aims at establishing correspondences between graph elements, and is widely used in many computer vision tasks. Among recently proposed graph matching algorithms, those utilizing the path following strategy have attracted special research attentions due to their exhibition of state-of-the-art performances. However, the paths computed in these algorithms often contain singular points, which could hurt the matching performance if not dealt properly. To deal with this issue, we propose a novel path following strategy, named branching path following (BPF), to improve graph matching accuracy. In particular, we first propose a singular point detector by solving a KKT system, and then design a branch switching method to seek for better paths at singular points. Moreover, to reduce the computational burden of the BPF strategy, an adaptive path estimation (APE) strategy is integrated into BPF to accelerate the convergence of searching along each path. A new graph matching algorithm named ABPF-G is developed by applying APE and BPF to a recently proposed path following algorithm named GNCCP (Liu & Qiao 2014). Experimental results reveal how our approach consistently outperforms state-of-the-art algorithms for graph matching on five public benchmark datasets.

Trajectory association and clustering are two key problems in motion analysis. While association links the points of interest to form trajectories, clustering discovers motion patterns of these trajectories and group them into clusters. Despite mutually related, the two problems have been typically studied separately in the literature. In this letter, we formulate them as a unified optimization problem and take the advantage of high-order information to capture the interrelations for Motion Segmentation by Trajectory Association and Clustering (MSTAC). To solve this unified problem, we propose an alternating optimization strategy to improve the association and clustering in each iteration. Specifically, a tensor-based multidimensional assignment method with high-order motion context information is proposed for trajectory association; and a minimum cost multicut-based trajectory clustering method is introduced for trajectory clustering. While the association process provides incomplete trajectories to clustering, the clustering method presents high-order context information to improve the performance of association; and thus they benefit from each other. Experiments on the Hopkins 155 dataset and a realistic airport sequence demonstrate that the proposed MSTAC framework obtains high accuracy on both trajectory association and clustering.

As an important and challenging problem in computer vision, zero-shot learning (ZSL) aims at automatically recognizing the instances from unseen object classes without training data. To address this problem, ZSL is usually carried out in the following two aspects: 1) capturing the domain distribution connections between seen classes data and unseen classes data and 2) modeling the semantic interactions between the image feature space and the label embedding space. Motivated by these observations, we propose a bidirectional mapping-based semantic relationship modeling scheme that seeks for cross-modal knowledge transfer by simultaneously projecting the image features and label embeddings into a common latent space. Namely, we have a bidirectional connection relationship that takes place from the image feature space to the latent space as well as from the label embedding space to the latent space. To deal with the domain shift problem, we further present a transductive learning approach that formulates the class prediction problem in an iterative refining process, where the object classification capacity is progressively reinforced through bootstrapping-based model updating over highly reliable instances. Experimental results on four benchmark datasets (animal with attribute, Caltech-UCSD Bird2011, aPascal-aYahoo, and SUN) demonstrate the effectiveness of the proposed approach against the state-of-the-art approaches.

Intensity inhomogeneity is common in real-world images and inevitably leads to many difficulties for accurate image segmentation. Numerous level-set methods have been proposed to segment images with intensity inhomogeneity. However, most of these methods are based on linear approximation, such as locally weighted mean, which may cause problems when handling images with severe intensity inhomogeneities. In this paper, we view segmentation of such images as a nonconvex optimization problem, since the intensity variation in such an image follows a nonlinear distribution. Then, we propose a novel level-set method named local approximation of Taylor expansion (LATE), which is a nonlinear approximation method to solve the nonconvex optimization problem. In LATE, we use the statistical information of the local region as a fidelity term and the differentials of intensity inhomogeneity as an adjusting term to model the approximation function. In particular, since the first-order differential is represented by the variation degree of intensity inhomogeneity, LATE can improve the approximation quality and enhance the local intensity contrast of images with severe intensity inhomogeneity. Moreover, LATE solves the optimization of function fitting by relaxing the constraint condition. In addition, LATE can be viewed as a constraint relaxation of classical methods, such as the region-scalable fitting model and the local intensity clustering model. Finally, the level-set energy functional is constructed based on the Taylor expansion approximation. To validate the effectiveness of our method, we conduct thorough experiments on synthetic and real images. Experimental results show that the proposed method clearly outperforms other solutions in comparison.

Understanding traffic scene images taken from vehicle mounted cameras is important for high-level tasks, such as advanced driver assistance systems and autonomous driving. It is a challenging problem due to large variations under different weather or illumination conditions. In this paper, we tackle the problem of traffic scene understanding from a cross-domain perspective. We attempt to understand the traffic scene from images taken from the same location but under different weather or illumination conditions (e.g., understanding the same traffic scene from images on a rainy night with the help of images taken on a sunny day). To this end, we propose a dense correspondence-based transfer learning (DCTL) approach, which consists of three main steps: 1) extracting deep representations of traffic scene images via a fine-tuned convolutional neural network; 2) constructing compact and effective representations via cross-domain metric learning and subspace alignment for cross-domain retrieval; and 3) transferring the annotations from the retrieved best matching image to the test image based on cross-domain dense correspondences and a probabilistic Markov random field. To verify the effectiveness of our DCTL approach, we conduct extensive experiments on a challenging data set, which contains 1828 images from six weather or illumination conditions.

Image context in image is crucial for improving scene labeling. While the existing methods only exploit local context generated from a small surrounding area of an image patch or a pixel, the long-range and global contextual information is often ignored. To handle this issue, we propose a novel approach for scene labeling by multi-level contextual recurrent neural networks (RNNs). We encode three kinds of contextual cues, viz., local context, global context, and image topic context in structural RNNs to model long-range local and global dependencies in an image. In this way, our method is able to “see” the image in terms of both long-range local and holistic views, and make a more reliable inference for image labeling. Besides, we integrate the proposed contextual RNNs into hierarchical convolutional neural networks, and exploit dependence relationships at multiple levels to provide rich spatial and semantic information. Moreover, we adopt an attention model to effectively merge multiple levels and show that it outperforms average- or max-pooling fusion strategies. Extensive experiments demonstrate that the proposed approach achieves improved results on the CamVid, KITTI, SiftFlow, Stanford Background, and Cityscapes data sets.

We propose a set packing (SP) framework for joint headlight pairing and vehicle tracking. Given headlight detections, traditional nighttime vehicle tracking methods usually first pair headlights and then track these pairs. However, the poor photometric condition often introduces tremendous noises in headlight detection and pairing, which leads to unrecoverable errors for vehicle tracking. To overcome the challenge, we propose to jointly model these two tasks in a weighted SP framework. Specifically, a graph is built which takes candidate pair track hypotheses as nodes and encodes in edges both the disjoint constraints for tracking and the no-sharing-headlight constraints for pairing. Solving a weighted SP problem on such a graph produces vehicle trajectories, and facilitates pairing with temporal context and in turn produces high quality vehicle trajectories. The solution, however, raises the issue of unmanageable graph scale since the number of track hypotheses grows exponentially over time. To address this issue, pruning strategies are developed to solve the joint model efficiently. The proposed system is evaluated on two traffic data sets, including videos under various challenging conditions. Both quantitative and qualitative results show that our method outperforms other tested methods, both in nighttime vehicle tracking and in multi-target tracking, confirming the benefits of jointly modeling the two tasks.

Being intensively studied, visual tracking has seen great recent advances in either speed (e.g., with correlation filters) or accuracy (e.g., with deep features). Real-time and high accuracy tracking algorithms, however, remain scarce. In this paper we study the problem from a new perspective and present a novel parallel tracking and verifying (PTAV) framework, by taking advantage of the ubiquity of multithread techniques and borrowing from the success of parallel tracking and mapping in visual SLAM. Our PTAV framework typically consists of two components, a tracker T and a verifier V, working in parallel on two separate threads. The tracker T aims to provide a super real-time tracking inference and is expected to perform well most of the time; by contrast, the verifier V checks the tracking results and corrects T when needed. The key innovation is that, V does not work on every frame but only upon the requests from T; on the other end, T may adjust the tracking according to the feedback from V. With such collaboration, PTAV enjoys both the high efficiency provided by T and the strong discriminative power by V. In our extensive experiments on popular benchmarks including OTB2013, OTB2015, TC128 and UAV20L, PTAV achieves the best tracking accuracy among all real-time trackers, and in fact performs even better than many deep learning based solutions. Moreover, as a general framework, PTAV is very flexible and has great rooms for improvement and generalization.

In this paper we propose a new salient object detection method via structured label prediction. By learning appearance features in rectangular regions, our structural region representation encodes the local saliency distribution with a matrix of binary labels. We show that the linear combination of structured labels can well model the saliency distribution in local regions. Representing region saliency with structured labels has two advantages: 1) it connects the label assignment of all enclosed pixels, which produces a smooth saliency prediction; and 2) regularshaped nature of structured labels enables well definition of traditional cues such as regional properties and center surround contrast, and these cues help to build meaningful and informative saliency measures. To measure the consistency between a structured label and the corresponding saliency distribution, we further propose an adaptive label ranking algorithm using proposals that are generated by a CNN model. Finally, we introduce a K-NN enhanced graph representation for saliency propagation, which is more favorable for our task than the widely-used adjacent-graph-based ones. Experimental results demonstrate the effectiveness of our proposed method on six popular benchmarks compared with state-of-the-art approaches.

Tracking for planar objects is an important issue to vision-based robotic applications. In direct visual tracking (DVT) methods, the similarity between two images is often measured through the sum of squared differences (SSD) especially with the efficient second-order minimization (ESM) due to its simplicity and efficiency. However, SSD-based ESM is not robust to illumination changes since it is usually built upon the brightness constancy assumption. Contrast to image brightness, gradient orientations (GO) are invariant to both linear and non-linear illumination changes as verified in practice. Based on GO, we propose an illumination insensitive ESM method for planar object tracking in this paper. In order to introduce GO into the ESM, we generalized the original ESM formulas for multi-dimensional features. In addition, a denoising method based on the Perona-Malik function and a mask image were suggested to improve GO's robustness against image noise and low texture. Our experimental results on dataset for planar objects with illumination changes and a benchmark dataset confirm the proposed method is robust to illumination variations and capable to deal with the general tracking challenges.

The diversity of base learners is of utmost importance to a good ensemble. This paper defines a novel measurement of diversity, termed as exclusivity. With the designed exclusivity, we further propose an ensemble SVM classifier, namely Exclusivity Regularized Machine (ExRM), to jointly suppress the training error of ensemble and enhance the diversity between bases. Moreover, an Augmented Lagrange Multiplier based algorithm is customized to effectively and efficiently seek the optimal solution of ExRM. Theoretical analysis on convergence, global optimality and linear complexity of the proposed algorithm, as well as experiments are provided to reveal the efficacy of our method and show its superiority over state-of-the-arts in terms of accuracy and efficiency.

Existing saliency detection approaches use images as inputs and are sensitive to foreground/background similarities, complex background textures, and occlusions. We explore the problem of using light fields as input for saliency detection. Our technique is enabled by the availability of commercial plenoptic cameras that capture the light field of a scene in a single shot. We show that the unique refocusing capability of light fields provides useful focusness, depths, and objectness cues. We further develop a new saliency detection algorithm tailored for light fields. To validate our approach, we acquire a light field database of a range of indoor and outdoor scenes and generate the ground truth saliency map. Experiments show that our saliency detection scheme can robustly handle challenging scenarios such as similar foreground and background, cluttered background, complex occlusions, etc., and achieve high accuracy and robustness.

Low-rank recovery models have shown potential for salient object detection, where a matrix is decomposed into a low-rank matrix representing image background and a sparse matrix identifying salient objects. Two deficiencies, however, still exist. First, previous work typically assumes the elements in the sparse matrix are mutually independent, ignoring the spatial and pattern relations of image regions. Second, when the low-rank and sparse matrices are relatively coherent, e.g., when there are similarities between the salient objects and background or when the background is complicated, it is difficult for previous models to disentangle them. To address these problems, we propose a novel structured matrix decomposition model with two structural regularizations: (1) a tree-structured sparsity-inducing regularization that captures the image structure and enforces patches from the same object to have similar saliency values, and (2) a Laplacian regularization that enlarges the gaps between salient objects and the background in feature space. Furthermore, high-level priors are integrated to guide the matrix decomposition and boost the detection. We evaluate our model for salient object detection on five challenging datasets including single object, multiple objects and complex scene images, and show competitive results as compared with 24 state-of-the-art methods in terms of seven performance metrics.

Skin detection from images, typically used as a preprocessing step, has a wide range of applications such as dermatology diagnostics, human computer interaction designs, and etc. It is a challenging problem due to many factors such as variation in pigment melanin, uneven illumination, and differences in ethnicity geographics. Besides, age and gender introduce additional difficulties to the detection process. It is hard to determine whether a single pixel is skin or nonskin without considering the context. An efficient traditional hand-engineered skin color detection algorithm requires extensive work by domain experts. Recently, deep learning algorithms, especially convolutional neural networks (CNNs), have achieved great success in pixel-wise labeling tasks. However, CNN-based architectures are not sufficient for modeling the relationship between pixels and their neighbors. In this letter, we integrate recurrent neural networks (RNNs) layers into the fully convolutional neural networks (FCNs), and develop an end-to-end network for human skin detection. In particular, FCN layers capture generic local features, while RNN layers model the semantic contextual dependencies in images. Experimental results on the COMPAQ and ECU skin datasets validate the effectiveness of the proposed approach, where RNN layers enhance the discriminative power of skin detection in complex background situations.

When one captures images in low-light conditions, the images often suffer from low visibility. Besides degrading the visual aesthetics of images, this poor quality may also significantly degenerate the performance of many computer vision and multimedia algorithms that are primarily designed for high-quality inputs. In this paper, we propose a simple yet effective low-light image enhancement (LIME) method. More concretely, the illumination of each pixel is first estimated individually by finding the maximum value in R, G, and B channels. Furthermore, we refine the initial illumination map by imposing a structure prior on it, as the final illumination map. Having the well-constructed illumination map, the enhancement can be achieved accordingly. Experiments on a number of challenging low-light images are present to reveal the efficacy of our LIME and show its superiority over several state-of-the-arts in terms of enhancement quality and efficiency.

Direction information serves as one of the most important features for palmprint recognition. In the past decade, many effective direction representation (DR)-based methods have been proposed and achieved promising recognition performance. However, due to an incomplete understanding for DR, these methods only extract DR in one direction level and one scale. Hence, they did not fully utilize all potentials of DR. In addition, most researchers only focused on the DR extraction in spatial coding domain, and rarely considered the methods in frequency domain. In this paper, we propose a general framework for DR-based method named complete DR (CDR), which reveals DR by a comprehensive and complete way. Different from traditional methods, CDR emphasizes the use of direction information with strategies of multi-scale, multi-direction level, multi-region, as well as feature selection or learning. This way, CDR subsumes previous methods as special cases. Moreover, thanks to its new insight, CDR can guide the design of new DR-based methods toward better performance. Motived this way, we propose a novel palmprint recognition algorithm in frequency domain. First, we extract CDR using multi-scale modified finite radon transformation. Then, an effective correlation filter, namely, band-limited phase-only correlation, is explored for pattern matching. To remove feature redundancy, the sequential forward selection method is used to select a small number of CDR images. Finally, the matching scores obtained from different selected features are integrated using score-level-fusion. Experiments demonstrate that our method can achieve better recognition accuracy than the other state-of-the-art methods. More importantly, it has fast matching speed, making it quite suitable for the large-scale identification applications.

Graph matching plays an important role in many fields in computer vision. It is a well-known general NP-hard problem and has been investigated for decades. Among the large amount of algorithms for graph matching, the algorithms utilizing the path following strategy exhibited state-of-art performances. However, the main drawback of this category of algorithms lies in their high computational burden. In this paper, we propose a novel path following strategy for graph matching aiming to improve its computation efficiency. We first propose a path estimation method to reduce the computational cost at each iteration, and subsequently a method of adaptive step length to accelerate the convergence. The proposed approach is able to be integrated into all the algorithms that utilize the path following strategy. To validate our approach, we compare our approach with several recently proposed graph matching algorithms on three benchmark image datasets. Experimental results show that, our approach improves significantly the computation efficiency of the original algorithms, and offers similar or better matching results.

Due to its wide range of applications, matching between two graphs has been extensively studied and remains an active topic. By contrast, it is still under-exploited on how to jointly match multiple graphs, partly due to its intrinsic combinatorial intractability. In this work, we address this challenging problem in a principled way under the rank-1 tensor approximation framework. In particular, we formulate multi-graph matching as a combinational optimization problem with two main ingredients: unary matching over graph vertices and structure matching over graph edges, both of which across multiple graphs. Then we propose an efficient power iteration solution for the resulting NP-hard optimization problem. The proposed algorithm has several advantages: 1) the intrinsic matching consistency across multiple graphs based on the high-order tensor optimization, 2) the free employment of powerful high-order node affinity, 3) the flexible integration between various types of node affinities and edge/hyper-edge affinities. Experiments on diverse and challenging datasets validate the effectiveness of the proposed approach in comparison with state-of the-arts.

Recently sparse representation has been applied to visual tracking by modeling the target appearance using a sparse approximation over the template set. However, this approach is limited by the high computational cost of the ℓ1-norm minimization involved, which also impacts on the amount of particle samples that we can have. This paper introduces a basic constraint on the self-representation of the target set. The sparsity pattern in the self-representation allows us to recover the “sparse coefficients” of the candidate samples by some small-scale ℓ2-norm minimization; this results in a fast tracking algorithm. It also leads to a principled dictionary update mechanism which is crucial for good performance. Experiments on a recently released benchmark with 50 challenging video sequences show significant runtime efficiency and tracking accuracy achieved by the proposed algorithm.

In this paper, we propose a vision-based approach for roadside vegetation detection by superpixel matching with local context. Unlike previous detection methods which seek help from additional sensors such as lidar, our algorithm only requires an off-the-shelf camera. The proposed method contains two stages. In the first stage, a superpixel database is constructed by segmenting training images into superpixels, and each superpixel patch is represented with multiple features. After that, the appearance information of vegetation or non-vegetation is encoded in the superpixel database. In the second stage, vegetation detection in each testing image is achieved by superpixel matching. The test image is segmented into superpixels and the (vegetation) label cost of each superpixel is derived by comparing with the k-nearest neighbors in the superpixel database. Furthermore, we incorporate the local context information through the feedback to refine superpixel matching. Taking this context information into account, Markov Random Field (MRF) is utilized to further improve the classification accuracy. Besides, considering the stable layout of road scene images, we utilize spatial priors of road scene to guide vegetation classification. Experiments on real-world datasets demonstrate the promise of our method.

To exploit the reliable prior knowledge that the target object in tracking must be an object other than nonobject, in this letter, we propose to adapt objectness for visual object tracking. Instead of directly applying an existing objectness measure that is generic and handles various objects and environments, we adapt it to be compatible to the specific tracking sequence and object. More specifically, we use the newly proposed binarized normed gradient (BING) objectness as the base, and then train an object-adaptive objectness for each tracking task. The training is implemented by using an adaptive support vector machine that integrates information from the specific tracking target into the BING measure. We emphasize that the benefit of the proposed adaptive objectness, named ADOBING, is generic. To show this, we combine ADOBING with eight top performed trackers in recent evaluations. We run the ADOBING-enhanced trackers along with their base trackers on the CVPR2013 benchmark, and our methods consistently improve the base trackers both in overall performance and under all challenge factors. Noting that the way we integrate objectness in visual tracking is generic and straightforward, we expect even more improvement by using tracker-specific objectness.

Dynamic scene classification started drawing an increasing amount of research efforts recently. While existing arts mainly rely on low-level features, little work addresses the need of exploring the rich spatial layout information in dynamic scene. Motivated by the fact that dynamic scenes are characterized by both dynamic and static parts with spatial layout priors, we propose to use redundant spatial grouping of a large number of spatiotemporal patches, named scenelet, to represent a dynamic scene. Specifically, each scenelet is associated with a category-dependent scenelet model to encode the likelihood of a specific scene category. All scenelet models for a scene category are jointly learned to encode the spatial interactions and redundancies among them. Subsequently, a dynamic scene sequence is represented as a collection of category likelihoods estimated by these scenelet models. Such presentation effectively encodes the spatial layout prior together with associated semantic information, and can be used for classifying dynamic scenes in combination with a standard learning algorithm such as k -nearest neighbor or linear support vector machine. The effectiveness of our approach is clearly demonstrated using two dynamic scene benchmarks and a related application for violence video classification. In the nearest neighbor classification framework, for dynamic scene classification, our method outperforms previous state-of-the-arts on both Maryland "in the wild" dataset and "stabilized" dynamic scene dataset. For violence video classification on a benchmark dataset, our method achieves a promising classification rate of 87.08%, which significantly improves previous best result of 81.30%.

A key problem in salient object detection is how to effectively model the semantic properties of salient objects in a data-driven manner. In this paper, we propose a multi-task deep saliency model based on a fully convolutional neural network with global input (whole raw images) and global output (whole saliency maps). In principle, the proposed saliency model takes a data-driven strategy for encoding the underlying saliency prior information, and then sets up a multi-task learning scheme for exploring the intrinsic correlations between saliency detection and semantic image segmentation. Through collaborative feature learning from such two correlated tasks, the shared fully convolutional layers produce effective features for object perception. Moreover, it is capable of capturing the semantic information on salient objects across different levels using the fully convolutional layers, which investigate the feature-sharing properties of salient object detection with a great reduction of feature redundancy. Finally, we present a graph Laplacian regularized nonlinear regression model for saliency refinement. Experimental results demonstrate the effectiveness of our approach in comparison with the state-of-the-art approaches.

One of the most logical applications of face recognition for authentication is on mobile handset devices. However, face recognition still faces challenges in providing environment tolerance: being able to compensate for changes in light conditions within an environment where authentication is occurring, due to users carrying their mobile handset devices to different locations with varying and unpredictable sources of illumination. Existing face recognition systems operate by finding fiduciary points relative to the area of the entire face, which becomes their weakness when they are not used in applications where light conditions are fixed and controlled. This research investigates Local Binary Patterns (LBP), an image encoding technique whose origins lie in texture analysis, in order to overcome the problems faced by existing face recognition systems and provide tolerance to variable light conditions. This research aims to utilize LBP on modern mobile handset device hardware that is "off-the-shelf": utilizing only the most basic and widely available onboard imaging hardware and processing capability provided on mobile handset devices of the present day. We have performed rigorous experimentation with LBP both on large databases of images of human faces, as well as developing mobile handset software that was deployed to real users and tested in a field environment. Our experimentation indicates that LBP is capable of being used to develop face recognition systems that provide environment tolerance, potentially finding practical use as a component of mobile device authentication applications.

In this paper we present a novel framework for cross-age face verification (FV) by seeking help from its “competitor” named cross-face age verification (AV), i.e., deciding whether two face photos are taken at similar ages. While FV and AV share some common features, FV pursues age insensitivity and AV seeks age sensitivity. Such correlation suggests that AV may be used to guide feature selection in FV, i.e., by reducing the chance of choosing age sensitive features. Driven by this intuition, we propose to learn a solution for cross-age face verification by coordinating with a solution for age verification. Specifically, a joint additive model is devised to simultaneously handling both tasks, while encoding feature coordination by a competition regularization term. Then, an alternating greedy coordinate descent (AGCD) algorithm is developed to solve this joint model. As shown in our experiments, the algorithm effectively balances feature sharing and feature exclusion between the two tasks; and, for face verification, the algorithm effectively removes distracting features used in age verification. To evaluate the proposed algorithm, we conduct cross-age face verification experiments using two benchmark cross-age face datasets, FG-Net and MORPH. In all experiments, our algorithm achieves very promising results and outperforms all previously tested solutions.

In this paper, we propose a real-time 3D hand pose estimation algorithm using the randomized decision forest framework. Our algorithm takes a depth image as input and generates a set of skeletal joints as output. Previous decision forest-based methods often give labels to all points in a point cloud at a very early stage and vote for the joint locations. By contrast, our algorithm only tracks a set of more flexible virtual landmark points, named segmentation index points (SIPs), before reaching the final decision at a leaf node. Roughly speaking, a SIP represents the centroid of a subset of skeletal joints, which are to be located at the leaves of the branch expanded from the SIP. Inspired by recent latent regression forest-based hand pose estimation framework (Tang et al. 2014), we integrate SIP into the framework with several important improvements: First, we devise a new forest growing strategy, whose decision is made using a randomized feature guided by SIPs. Second, we speed-up the training procedure since only SIPs, not the skeletal joints, are estimated at non-leaf nodes. Third, the experimental results on public benchmark datasets show clearly the advantage of the proposed algorithm over previous state-of-the-art methods, and our algorithm runs at 55.5 fps on a normal CPU without parallelism.

Plane geometric figures (PGFs) are essential diagrams that regularly appear in mathematics documents. To understand PGFs profoundly and intuitively, it is necessary to decompose them into visual elements rather than traditional line segments. In this paper, we present a method for detection and analysis of overlapped triangles on the basis of dominance rank. Overlapped triangles lead to a large quantity of redundant sub-triangles or incident triangles. The most important and representative triangles must be detected and adopted to reduce redundancy. Diverse types of relationships among potential triangles present at least three obstacles: (1) how to precisely detect all potential triangles, (2) how to select predominant triangles, and (3) how to verify that the combination of these predominant triangles can reconstruct the original figure. Based on Gestalt theory, we propose a method for selecting predominant triangles, including selection of the main convex triangle and composition of interior auxiliary triangles. Experiments show that our algorithm can detect most of the potential triangles and can present reasonable decomposition solutions using only few predominant elements.

The images we encounter throughout our lives make different impressions on us: Some are remembered at first glance, while others are forgotten. This phenomenon is caused by the intrinsic memorability of images revealed by recent studies [5,6]. In this paper, we address the issue of automatically estimating the memorability of images by proposing a novel multi-view adaptive regression (MAR) model. The MAR model provides an effective mapping of visual features to memorability scores by taking advantage of robust feature selection and multiple feature integration. It consists of three major components: an adaptive loss function, an adaptive regularization and a multi-view modeling strategy. Moreover, we design an alternating direction method (ADM) optimization algorithm to solve the proposed objective function. Experimental results on the MIT benchmark dataset show the superiority of the proposed model compared with existing image memorability prediction methods.

In this paper, we study a novel problem of classifying covert photos, whose acquisition processes are intentionally concealed from the subjects being photographed. Covert photos are often privacy invasive and, if distributed over Internet, can cause serious consequences. Automatic identification of such photos, therefore, serves as an important initial step toward further privacy protection operations. The problem is, however, very challenging due to the large semantic similarity between covert and noncovert photos, the enormous diversity in the photographing process and environment of cover photos, and the difficulty to collect an effective data set for the study. Attacking these challenges, we make three consecutive contributions. First, we collect a large data set containing 2500 covert photos, each of them is verified rigorously and carefully. Second, we conduct a user study on how humans distinguish covert photos from noncovert ones. The user study not only provides an important evaluation baseline, but also suggests fusing heterogeneous information for an automatic solution. Our third contribution is a covert photo classification algorithm that fuses various image features and visual attributes in the multiple kernel learning framework. We evaluate the proposed approach on the collected data set in comparison with other modern image classifiers. The results show that our approach achieves an average classification rate (1-EER) of 0.8940, which significantly outperforms other competitors as well as human's performance.

While color information is known to provide rich discriminative clues for visual inference, most modern visual trackers limit themselves to the grayscale realm. Despite recent efforts to integrate color in tracking, there is a lack of comprehensive understanding of the role color information can play. In this paper, we attack this problem by conducting a systematic study from both the algorithm and benchmark perspectives. On the algorithm side, we comprehensively encode 10 chromatic models into 16 carefully selected state-of-the-art visual trackers. On the benchmark side, we compile a large set of 128 color sequences with ground truth and challenge factor annotations (e.g., occlusion). A thorough evaluation is conducted by running all the color-encoded trackers, together with two recently proposed color trackers. A further validation is conducted on an RGBD tracking benchmark. The results clearly show the benefit of encoding color information for tracking. We also perform detailed analysis on several issues, including the behavior of various combinations between color model and visual tracker, the degree of difficulty of each sequence for tracking, and how different challenge factors affect the tracking performance. We expect the study to provide the guidance, motivation, and benchmark for future work on encoding color in visual tracking.

Nighttime traffic surveillance is difficult due to insufficient and unstable appearance information and strong background interference. We present in this paper a robust nighttime vehicle detection system by detecting, tracking, and grouping headlights. First, we train AdaBoost classifiers for headlights detection to reduce false alarms caused by reflections. Second, to take full advantage of the complementary nature of grouping and tracking, we alternately optimize grouping and tracking. For grouping, motion features produced by tracking are used by headlights pairing. We use a maximal independent set framework for effective pairing, which is more robust than traditional pairing-by-rules methods. For tracking, context information provided by pairing is employed by multiple object tracking. The experiments on challenging datasets and quantitative evaluation show promising performance of our method.

Joint learning of similar tasks has been a popular trend in visual recognition and proven to be beneficial. Between-task similarity often provides useful cues, such as feature sharing, for learning visual classifiers. By contrast, the competition relationship between visual recognition tasks (e.g., content independent writer identification and handwriting recognition) remains largely under-explored. A key challenge in visual recognition is to select the most discriminating features and remove irrelevant features related to intra-class variations. With the help of auxiliary competing tasks, we can identify such features within a joint learning model exploiting the competition relationship.Motivated by this intuition, we propose a novel way to exploit competition relationship for solving visual recognition problems. Specifically, given a target task and its competing tasks, we jointly model them by a generalized additive regression model with a competition constraint. This constraint effectively discourages choosing of irrelevant features (weak learners) that support the auxiliary competing tasks. We name the proposed algorithm CompBoost. In our study, CompBoost is applied to two visual recognition applications: (1) content-independent writer identification from handwriting scripts by exploiting competing tasks of handwriting recognition, and (2) actor-independent facial expression recognition by exploiting competing tasks of face recognition. In both experiments our approach demonstrates promising performance gains by exploiting the between-task competition.

Robust tracking of deformable object like catheter or vascular structures in X-ray images is an important technique used in image guided medical interventions for effective motion compensation and dynamic multi-modality image fusion. Tracking of such anatomical structures and devices is very challenging due to large degrees of appearance changes, low visibility of X-ray images and the deformable nature of the underlying motion field as a result of complex 3D anatomical movements projected into 2D images. To address these issues, we propose a new deformable tracking method using the tensor-based algorithm with model propagation. Specifically, the deformable tracking is formulated as a multi-dimensional assignment problem which is solved by rank-1 l1tensor approximation. The model prior is propagated in the course of deformable tracking. Both the higher order information and the model prior provide powerful discriminative cues for reducing ambiguity arising from the complex background, and consequently improve the tracking robustness. To validate the proposed approach, we applied it to catheter and vascular structures tracking and tested on X-ray fluoroscopic sequences obtained from 17 clinical cases. The results show, both quantitatively and qualitatively, that our approach achieves a mean tracking error of 1.4 pixels for vascular structure and 1.3 pixels for catheter tracking.

Existing saliency detection approaches use images as inputs and are sensitive to foreground/background similarities, complex background textures, and occlusions. We explore the problem of using light fields as input for saliency detection. Our technique is enabled by the availability of commercial plenoptic cameras that capture the light field of a scene in a single shot. We show that the unique refocusing capability of light fields provides useful focusness, depths, and objectness cues. We further develop a new saliency detection algorithm tailored for light fields. To validate our approach, we acquire a light field database of a range of indoor and outdoor scenes and generate the ground truth saliency map. Experiments show that our saliency detection scheme can robustly handle challenging scenarios such as similar foreground and background, cluttered background, complex occlusions, etc, and achieve high accuracy and robustness.

Interactions between moving targets often provide discriminative clues for multiple target tracking (MTT), though many existing approaches ignore such interactions due to difficulty in effectively handling them. In this paper, we model interactions between neighbor targets by pair-wise motion context, and further encode such context into the global association optimization. To solve the resulting global non-convex maximization, we propose an effective and efficient power iteration framework. This solution enjoys two advantages for MTT: First, it allows us to combine the global energy accumulated from individual trajectories and the between-trajectory interaction energy into a united optimization, which can be solved by the proposed power iteration algorithm. Second, the framework is flexible to accommodate various types of pairwise context models and we in fact studied two different context models in this paper. For evaluation, we apply the proposed methods to four public datasets involving different challenging scenarios such as dense aerial borne traffic tracking, dense point set tracking, and semi-crowded pedestrian tracking. In all the experiments, our approaches demonstrate very promising results in comparison with state-of-the-art trackers.

Face de-identification, the process of preventing a person' identity from being connected with personal information, is an important privacy protection tool in multimedia data processing. With the advance of face detection algorithms, a natural solution is to blur or block facial regions in visual data so as to obscure identity information. Such solutions however often destroy privacy-insensitive information and hence limit the data utility, e.g., gender and age information. In this paper we address the de-identification problem by proposing a simple yet effective framework, named GARP-Face, that balances utility preservation in face deidentification. In particular, we use modern facial analysis technologies to determine the Gender, Age, and Race attributes of facial images, and Preserving these attributes by seeking corresponding representatives constructed through a gallery dataset. We evaluate the proposed approach using the MORPH dataset in comparison with several state-of-the-art face de-identification solutions. The results show that our method outperforms previous solutions in preserving data utility while achieving similar degree of privacy protection.

Large variations in image background may cause partial matching and normalization problems for histogram-based representations, i.e., the histograms of the same category may have bins which are significantly different, and normalization may produce large changes in the differences between corresponding bins. In this paper, we deal with this problem by using the ratios between bin values of histograms, rather than bin values' differences which are used in the traditional histogram distances. We propose a bin ratio-based histogram distance (BRD), which is an intra-cross-bin distance, in contrast with previous bin-to-bin distances and cross-bin distances. The BRD is robust to partial matching and histogram normalization, and captures correlations between bins with only a linear computational complexity. We combine the BRD with the ℓ1 histogram distance and the χ(2) histogram distance to generate the ℓ1 BRD and the χ(2) BRD, respectively. These combinations exploit and benefit from the robustness of the BRD under partial matching and the robustness of the ℓ1 and χ(2) distances to small noise. We propose a method for assessing the robustness of histogram distances to partial matching. The BRDs and logistic regression-based histogram fusion are applied to image classification. The experimental results on synthetic data sets show the robustness of the BRDs to partial matching, and the experiments on seven benchmark data sets demonstrate promising results of the BRDs for image classification.

In visual tracking, holistic and part-based representations are both popular choices to model target appearance. The former is known for great efficiency and convenience, while the latter for robustness against local appearance or shape variations. Based on nonnegative matrix factorization (NMF), we propose a novel visual tracker that takes advantage of both groups. The idea is to model the target appearance by a nonnegative combination of nonnegative components learned from examples observed in previous frames. To adjust NMF to the tracking context, we include sparsity and smoothness constraints in addition to the nonnegativity one. Furthermore, an online iterative learning algorithm, together with a proof of convergence, is proposed for efficient model updating. Putting these ingredients together with a particle filter framework, the proposed tracker, constrained online nonnegative matrix factorization (CONMF), achieves robustness to challenging appearance variations and nontrivial deformations while running in real time. We evaluate the proposed tracker on various benchmark sequences containing targets undergoing large variations in scale, pose, or illumination. The robustness and efficiency of CONMF is validated in comparison with several state-of-the-art trackers.

In this paper, we propose saliency driven image multiscale nonlinear diffusion filtering. The resulting scale space in general preserves or even enhances semantically important structures such as edges, lines, or flow-like structures in the foreground, and inhibits and smoothes clutter in the background. The image is classified using multiscale information fusion based on the original image, the image at the final scale at which the diffusion process converges, and the image at a midscale. Our algorithm emphasizes the foreground features, which are important for image classification. The background image regions, whether considered as contexts of the foreground or noise to the foreground, can be globally handled by fusing information from different scales. Experimental tests of the effectiveness of the multiscale space for the image classification are conducted on the following publicly available datasets: 1) the PASCAL 2005 dataset; 2) the Oxford 102 flowers dataset; and 3) the Oxford 17 flowers dataset, with high classification rates.

In this paper, we propose two novel shape descriptors, angular pattern (AP) and binary angular pattern (BAP), and a multiscale integration of them for shape retrieval. Both AP and BAP are intrinsically invariant to scale and rotation. More importantly, being global shape descriptors, the proposed shape descriptors are computationally very efficient, while possessing similar discriminability as state-of-the-art local descriptors. As a result, the proposed approach is attractive for real world shape retrieval applications. The experiments on the widely used MPEG-7 and TARI-1000 data sets demonstrate the effectiveness of the proposed method in comparison with existing methods.

In this paper, we propose using high-level action units to represent human actions in videos and, based on such units, a novel sparse model is developed for human action recognition. There are three interconnected components in our approach. First, we propose a new context-aware spatial-temporal descriptor, named locally weighted word context, to improve the discriminability of the traditionally used local spatial-temporal descriptors. Second, from the statistics of the context-aware descriptors, we learn action units using the graph regularized nonnegative matrix factorization, which leads to a part-based representation and encodes the geometrical information. These units effectively bridge the semantic gap in action recognition. Third, we propose a sparse model based on a joint l2,1-norm to preserve the representative items and suppress noise in the action units. Intuitively, when learning the dictionary for action representation, the sparse model captures the fact that actions from the same class share similar units. The proposed approach is evaluated on several publicly available data sets. The experimental results and analysis clearly demonstrate the effectiveness of the proposed approach.

In this paper, we present a new geometric-temporal representation for visual action recognition based on local spatio-temporal features. First, we propose a modified covariance descriptor under the log-Euclidean Riemannian metric to represent the spatio-temporal cuboids detected in the video sequences. Compared with previously proposed covariance descriptors, our descriptor can be measured and clustered in Euclidian space. Second, to capture the geometric-temporal contextual information, we construct a directional pyramid co-occurrence matrix (DPCM) to describe the spatio-temporal distribution of the vector-quantized local feature descriptors extracted from a video. DPCM characterizes the co-occurrence statistics of local features as well as the spatio-temporal positional relationships among the concurrent features. These statistics provide strong descriptive power for action recognition. To use DPCM for action recognition, we propose a directional pyramid co-occurrence matching kernel to measure the similarity of videos. The proposed method achieves the state-of-the-art performance and improves on the recognition performance of the bag-of-visual-words (BOVWs) models by a large margin on six public data sets. For example, on the KTH data set, it achieves 98.78% accuracy while the BOVW approach only achieves 88.06%. On both Weizmann and UCF CIL data sets, the highest possible accuracy of 100% is achieved.

In this paper we formulate multi-target tracking (MTT) as a rank-1 tensor approximation problem and propose an ℓ1norm tensor power iteration solution. In particular, a high order tensor is constructed based on trajectories in the time window, with each tensor element as the affinity of the corresponding trajectory candidate. The local assignment variables are the ℓ1normalized vectors, which are used to approximate the rank-1 tensor. Our approach provides a flexible and effective formulation where both pairwise and high-order association energies can be used expediently. We also show the close relation between our formulation and the multi-dimensional assignment (MDA) model. To solve the optimization in the rank-1 tensor approximation, we propose an algorithm that iteratively powers the intermediate solution followed by an ℓ1normalization. Aside from effectively capturing high-order motion information, the proposed solver runs efficiently with proved convergence. The experimental validations are conducted on two challenging datasets and our method demonstrates promising performances on both.

Spatio-temporal interest points serve as an elementary building block in many modern action recognition algorithms, and most of them exploit the local spatio-temporal volume features using a Bag of Visual Words (BOVW) representation. Such representation, however, ignores potentially valuable information about the global spatio-temporal distribution of interest points. In this paper, we propose a new global feature to capture the detailed geometrical distribution of interest points. It is calculated by using the R transform which is defined as an extended 3D discrete Radon transform, followed by applying a two-directional two-dimensional principal component analysis. Such R feature captures the geometrical information of the interest points and keeps invariant to geometry transformation and robust to noise. In addition, we propose a new fusion strategy to combine the R feature with the BOVW representation for further improving recognition accuracy. We utilize a context-aware fusion method to capture both the pairwise similarities and higher-order contextual interactions of the videos. Experimental results on several publicly available datasets demonstrate the effectiveness of the proposed approach for action recognition.

Tracking multiple vehicles in wide area traffic scenes is challenging due to high target density, severe similar target ambiguity, and low frame rate. In this paper, we propose a novel spatio-temporal context model, named maximum consistency context (MCC), to leverage the discriminative power and robustness in the scenario. For a candidate association, its MCC is defined as the most consistent association in its neighborhood. Such a maximum selection picks the reliable neighborhood context information while filtering out noisy distraction. We tested the proposed context modeling on multi-target tracking using three challenging wide area motion sequences. Both quantitative and qualitative results show clearly the effectiveness of MCC, in comparison with algorithms that use no context and standard spatial context respectively.

The goal of saliency detection is to locate important pixels or regions in an image which attract humans' visual attention the most. This is a fundamental task whose output may serve as the basis for further computer vision tasks like segmentation, resizing, tracking and so forth. In this paper we propose a novel salient region detection algorithm by integrating three important visual cues namely uniqueness, focus ness and objectness (UFO). In particular, uniqueness captures the appearance-derived visual contrast, focus ness reflects the fact that salient regions are often photographed in focus, and objectness helps keep completeness of detected salient regions. While uniqueness has been used for saliency detection for long, it is new to integrate focus ness and objectness for this purpose. In fact, focus ness and objectness both provide important saliency information complementary of uniqueness. In our experiments using public benchmark datasets, we show that, even with a simple pixel level combination of the three components, the proposed approach yields significant improvement compared with previously reported methods.

Evaluating visual tracking algorithms, or trackers for short, is of great importance in computer vision. However, it is hard to fairly compare trackers due to many parameters need to be tuned in the experimental configurations. On the other hand, when introducing a new tracker, a recent trend is to validate it by comparing it with several existing ones. Such an evaluation may have subjective biases towards the new tracker which typically performs the best. This is mainly due to the difficulty to optimally tune all its competitors and sometimes the selected testing sequences. By contrast, little subjective bias exists towards the second best ones in the contest. This observation inspires us with a novel perspective towards inhibiting subjective bias in evaluating trackers by analyzing the results between the second bests. In particular, we first collect all tracking papers published in major computer vision venues in recent years. From these papers, after filtering out potential biases in various aspects, we create a dataset containing many records of comparison results between various visual trackers. Using these records, we derive performance rankings of the involved trackers by four different methods. The first two methods model the dataset as a graph and then derive the rankings over the graph, one by a rank aggregation algorithm and the other by a PageRank-like solution. The other two methods take the records as generated from sports contests and adopt widely used Elo's and Glicko's rating systems to derive the rankings. The experimental results are presented and may serve as a reference for related research.

Light fields are image-based representations that use densely sampled rays as a scene description. In this paper, we explore geometric structures of 3D lines in ray space for improving light field triangulation and stereo matching. The triangulation problem aims to fill in the ray space with continuous and non-overlapping simplices anchored at sampled points (rays). Such a triangulation provides a piecewise-linear interpolant useful for light field super-resolution. We show that the light field space is largely bilinear due to 3D line segments in the scene, and direct triangulation of these bilinear subspaces leads to large errors. We instead present a simple but effective algorithm to first map bilinear subspaces to line constraints and then apply Constrained Delaunay Triangulation (CDT). Based on our analysis, we further develop a novel line-assisted graph-cut (LAGC) algorithm that effectively encodes 3D line constraints into light field stereo matching. Experiments on synthetic and real data show that both our triangulation and LAGC algorithms outperform state-of-the-art solutions in accuracy and visual quality.

Digital plane geometric figures (PGFs) are important resources of digital education, especially in mathematical pedagogy. The related applications, such as recognition and retrieval system on geometric figure images, have not been fully exploited. Special quadrangles, including rectangles, parallelograms, and trapezoids, are important components of PGFs, and extracting these special quadrangles is a prerequisite task. In this paper, we focus on the detection of overlapped quadrangles in PGFs using the proposed one-pass detection algorithm based on the geometric symmetrical property of involved shapes. We also introduce a method to refine the detected geometric primitives by significance analysis in order to optimize the shape descriptor. Experimental results show that the proposed method for special quadrangle detection is accurate and efficient, and is ready for further use in PGF retrieval systems.

In this paper, we propose a new texture descriptor for both static and dynamic textures. The new descriptor is built on the wavelet-based spatial-frequency analysis of two complementary wavelet pyramids: standard multiscale and wavelet leader. These wavelet pyramids essentially capture the local texture responses in multiple high-pass channels in a multiscale and multiorientation fashion, in which there exists a strong power-law relationship for natural images. Such a power-law relationship is characterized by the so-called multifractal analysis. In addition, two more techniques, scale normalization and multiorientation image averaging, are introduced to further improve the robustness of the proposed descriptor. Combining these techniques, the proposed descriptor enjoys both high discriminative power and robustness against many environmental changes. We apply the descriptor for classifying both static and dynamic textures. Our method has demonstrated excellent performance in comparison with the state-of-the-art approaches in several public benchmark datasets.

Recently, sparse representation has been applied to visual tracking to find the target with the minimum reconstruction error from a target template subspace. Though effective, these L1 trackers require high computational costs due to numerous calculations for l1 minimization. In addition, the inherent occlusion insensitivity of the l1 minimization has not been fully characterized. In this paper, we propose an efficient L1 tracker, named bounded particle resampling (BPR)-L1 tracker, with a minimum error bound and occlusion detection. First, the minimum error bound is calculated from a linear least squares equation and serves as a guide for particle resampling in a particle filter (PF) framework. Most of the insignificant samples are removed before solving the computationally expensive l1 minimization in a two-step testing. The first step, named τ testing, compares the sample observation likelihood to an ordered set of thresholds to remove insignificant samples without loss of resampling precision. The second step, named max testing, identifies the largest sample probability relative to the target to further remove insignificant samples without altering the tracking result of the current frame. Though sacrificing minimal precision during resampling, max testing achieves significant speed up on top of τ testing. The BPR-L1 technique can also be beneficial to other trackers that have minimum error bounds in a PF framework, especially for trackers based on sparse representations. After the error-bound calculation, BPR-L1 performs occlusion detection by investigating the trivial coefficients in the l1 minimization. These coefficients, by design, contain rich information about image corruptions, including occlusion. Detected occlusions are then used to enhance the template updating. For evaluation, we conduct experiments on three video applications: biometrics (head movement, hand holding object, singers on stage), pedestrians (urban travel, hallway monitoring), and cars in traffic (wide area motion imagery, ground-mounted perspectives). The proposed BPR-L1 method demonstrates an excellent performance as compared with nine state-of-the-art trackers on eleven challenging benchmark sequences.

Recently sparse representation has been applied to visual tracker by modeling the target appearance using a sparse approximation over a template set, which leads to the so-called L1 trackers as it needs to solve an ℓ1norm related minimization problem for many times. While these L1 trackers showed impressive tracking accuracies, they are very computationally demanding and the speed bottleneck is the solver to ℓ1norm minimizations. This paper aims at developing an L1 tracker that not only runs in real time but also enjoys better robustness than other L1 trackers. In our proposed L1 tracker, a new ℓ1norm related minimization model is proposed to improve the tracking accuracy by adding an ℓ1norm regularization on the coefficients associated with the trivial templates. Moreover, based on the accelerated proximal gradient approach, a very fast numerical solver is developed to solve the resulting ℓ1norm related minimization problem with guaranteed quadratic convergence. The great running time efficiency and tracking accuracy of the proposed tracker is validated with a comprehensive evaluation involving eight challenging sequences and five alternative state-of-the-art trackers.

The advances in image acquisition techniques make recording images never easier and brings a great convenience to our daily life. It raises at the same time the issue of privacy protection in the photographs. One particular problem addressed in this paper is about covert photographs, which are taken secretly and often violate the subjects' willingness. We study the task of automatic covert photograph classification, which can be used to help inhibiting distribution of such images (e.g., Internet image filtering). By carefully collecting and investigating a large covert vs. non-covert photographs dataset, we observed that there are many features (e.g., degree of blur) that seem to be correlated with covert photographs, but counter examples always exist. In addition, we observed that image visual attributes (e.g., photo composition) play an important role in distinguishing covert photographs. These observations motivate us to fuse both low level images statistics and middle level attribute features for classifying covert images. In particular, we propose a solution using multiple kernel learning to combine 10 different image features and 31 image attributes. We evaluated thoroughly the proposed approach together with many different solutions including some state-of-the-art image classifiers. The effectiveness of the proposed solution is clearly demonstrated in the results. Furthermore, as the first study to this problem, we expect our study to motivate further research investigations.

In this paper we study the problem of online aligning a newly arrived image to previously well-aligned images. Inspired by recent advances in batch image alignment using low rank decomposition [16], we treat the newly arrived image, after alignment, as being linearly and sparsely reconstructed by the well-aligned ones. The task is accomplished by a sequence of convex optimization that minimizes the l\-norm. After that, online basis updating is pursued in two different ways: (1) a two-stage incremental alignment for joint registration of a large image dataset which is known a prior, and (2) a greedy online alignment of dynamically increasing image sequences, such as in the tracking scenario. In (1), we first sequentially collect basis images that are easily aligned by checking their reconstruction residuals, followed by the second stage where all images are re-aligned one-by-one using the collected basis set. In (2), during the tracking process, we dynamically enrich the image basis set by the new target if it significantly distinguishes itself from existing basis images. While inheriting the benefits of sparsity, our method enjoys the great time efficiency and therefore be capable of dealing with large image set and real time tasks such as visual tracking. The efficacy of the proposed online robust alignment algorithm is verified with extensive experiments on image set alignment and visual tracking, in reference with state-of-the-art methods.

BACKGROUND: The relationships between the gene functional similarity and gene expression profile, and between gene function annotation and gene sequence have been studied extensively. However, not much work has considered the connection between gene functions and location of a gene's expression in the mammalian tissues. On the other hand, although unsupervised learning methods have been commonly used in functional genomics, supervised learning cannot be directly applied to a set of normal genes without having a target (class) attribute. RESULTS: Here, we propose a supervised learning methodology to predict pair-wise gene functional similarity from multiplex gene expression maps that provide information about the location of gene expression. The features are extracted from expression maps and the labels denote the functional similarities of pairs of genes. We make use of wavelet features, original expression values, difference and average values of neighboring voxels and other features to perform boosting analysis. The experimental results show that with increasing similarities of gene expression maps, the functional similarities are increased too. The model predicts the functional similarities between genes to a certain degree. The weights of the features in the model indicate the features that are more significant for this prediction. CONCLUSIONS: By considering pairs of genes, we propose a supervised learning methodology to predict pair-wise gene functional similarity from multiplex gene expression maps. We also explore the relationship between similarities of gene maps and gene functions. By using AdaBoost coupled with our proposed weak classifier we analyze a large-scale gene expression dataset and predict gene functional similarities. We also detect the most significant single voxels and pairs of neighboring voxels and visualize them in the expression map image of a mouse brain. This work is very important for predicting functions of unknown genes. It also has broader applicability since the methodology can be applied to analyze any large-scale dataset without a target attribute and is not restricted to gene expressions.

In this brief, we propose a novel contour-based shape descriptor, called the multiscale distance matrix, to capture the shape geometry while being invariant to translation, rotation, scaling, and bilateral symmetry. The descriptor is further combined with a dimensionality reduction to improve its discriminative power. The proposed method avoids the time-consuming pointwise matching encountered in most of the previously used shape recognition algorithms. It is therefore fast and suitable for real-time applications. We applied the proposed method to the task of plan leaf recognition with experiments on two data sets, the Swedish Leaf data set and the ICL Leaf data set. The experimental results clearly demonstrate the effectiveness and efficiency of the proposed descriptor.

The recently proposed covariance region descriptor has been proven robust and versatile for a modest computational cost. The covariance matrix enables efficient fusion of different types of features, where the spatial and statistical properties, as well as their correlation, are characterized. The similarity between two covariance descriptors is measured on Riemannian manifolds. Based on the same metric but with a probabilistic framework, we propose a novel tracking approach on Riemannian manifolds with a novel incremental covariance tensor learning (ICTL). To address the appearance variations, ICTL incrementally learns a low-dimensional covariance tensor representation and efficiently adapts online to appearance changes of the target with only O(1) computational complexity, resulting in a real-time performance. The covariance-based representation and the ICTL are then combined with the particle filter framework to allow better handling of background clutter, as well as the temporary occlusions. We test the proposed probabilistic ICTL tracker on numerous benchmark sequences involving different types of challenges including occlusions and variations in illumination, scale, and pose. The proposed approach demonstrates excellent real-time performance, both qualitatively and quantitatively, in comparison with several previously proposed trackers.

Recently, sparse representation has been applied to visual tracking to find the target with the minimum reconstruction error from the target template subspace. Though effective, these L1 trackers require high computational costs due to numerous calculations for ℓ1minimization. In addition, the inherent occlusion insensitivity of the ℓ1minimization has not been fully utilized. In this paper, we propose an efficient L1 tracker with minimum error bound and occlusion detection which we call Bounded Particle Resampling (BPR)-L1 tracker. First, the minimum error bound is quickly calculated from a linear least squares equation, and serves as a guide for particle resampling in a particle filter framework. Without loss of precision during resampling, most insignificant samples are removed before solving the computationally expensive ℓ1minimization function. The BPR technique enables us to speed up the L1 tracker without sacrificing accuracy. Second, we perform occlusion detection by investigating the trivial coefficients in the ℓ1minimization. These coefficients, by design, contain rich information about image corruptions including occlusion. Detected occlusions enhance the template updates to effectively reduce the drifting problem. The proposed method shows good performance as compared with several state-of-the-art trackers on challenging benchmark sequences.

Discriminative tracking has become popular tracking methods due to their descriptive power for foreground/background separation. Among these methods, online random forest is recently proposed and received a large amount of research attention due to its advantages such as efficiency and robust ness to noise, etc. However, the fact that only one kind of features is used limits the discriminative performance of this tracker. Additionally, the standard online forest tracker works only for a single target object. In this paper, we introduce a novel tracking method that integrates multiple cues capturing both geometric structures and edge-based shape information. Compared with the current online random forest based tracking algorithm, the proposed multi-cue tracker is more robust thanks to the complimentary information provided from these hybrid cues. Furthermore, the new tracker can track multiple targets as well as single target object. The effectiveness of the proposed tracker is validated using five public sequences.

Many image retargeting algorithms, despite aesthetically carving images smaller, pay limited attention to image browsing tasks where tiny thumbnails are presented. When applying traditional retargeting methods for generating thumbnails, several important issues frequently arise, including thumbnail scales, object completeness and local structure smoothness. To address these issues, we propose a novel image retargeting algorithm, Scale and Object Aware Retargeting (SOAR), which has four components: (1) a scale dependent saliency map to integrate size information of thumbnails, (2) objectness (Alexe et al. 2010) for preserving object completeness, (3) a cyclic seam carving algorithm to guide continuous retarget warping, and (4) a thin-plate-spline (TPS) retarget warping algorithm that champions local structure smoothness. The effectiveness of the proposed algorithm is evaluated both quantitatively and qualitatively. The quantitative evaluation is conducted through an image browsing user study to measure the effectiveness of different thumbnail generating algorithms, followed by the ANOVA analysis. The qualitative study is performed on the RetargetMe benchmark dataset. In both studies, SOAR generates very promising performance, in comparison with state-of-the-art retargeting algorithms.

Visual tracking plays an important role in many computer vision tasks. A common assumption in previous methods is that the video frames are blur free. In reality, motion blurs are pervasive in the real videos. In this paper we present a novel BLUr-driven Tracker (BLUT) framework for tracking motion-blurred targets. BLUT actively uses the information from blurs without performing debluring. Specifically, we integrate the tracking problem with the motion-from-blur problem under a unified sparse approximation framework. We further use the motion information inferred by blurs to guide the sampling process in the particle filter based tracking. To evaluate our method, we have collected a large number of video sequences with significant motion blurs and compared BLUT with state-of-the-art trackers. Experimental results show that, while many previous methods are sensitive to motion blurs, BLUT can robustly and reliably track severely blurred targets.

In this paper, we developed a novel tool called dynamic fractal analysis for dynamic texture (DT) classification, which not only provides a rich description of DT but also has strong robustness to environmental changes. The resulting dynamic fractal spectrum (DFS) for DT sequences consists of two components: One is the volumetric dynamic fractal spectrum component (V-DFS) that captures the stochastic self-similarities of DT sequences as 3D volume datasets; the other is the multi-slice dynamic fractal spectrum component (S-DFS) that encodes fractal structures of DT sequences on 2D slices along different views of the 3D volume. Various types of measures of DT sequences are collected in our approach to analyze DT sequences from different perspectives. The experimental evaluation is conducted on three widely used benchmark datasets. In all the experiments, our method demonstrated excellent performance in comparison with state-of-the-art approaches.

Advances in imaging devices and web technologies have brought dramatic improvements in collecting, storing, and sharing images. The leakage of privacy information in the process becomes an important issue that has started drawing attention from both academia and industry. In this work, we study the problem of privacy preserving with focus on license plate number protecting in imagery. Specifically, we present a novel method for de-identifying license plate images with the least degradation in image visual quality for privacy protection. Unlike previous de-identification methods that pay little attention to the image quality preservation, our method, named inhomogeneous principal component blur (IPCB), adaptively blurs different pixels of a license plate by taking into account the prior distribution of sensitive information. We tested the proposed method on a public dataset in comparison with several popular de-identification methods. The evaluation shows that our method successfully de-identified the privacy information with the least damage of image quality when compared with several other solutions.

In this paper, we propose a robust visual tracking method by casting tracking as a sparse approximation problem in a particle filter framework. In this framework, occlusion, noise, and other challenging issues are addressed seamlessly through a set of trivial templates. Specifically, to find the tracking target in a new frame, each target candidate is sparsely represented in the space spanned by target templates and trivial templates. The sparsity is achieved by solving an l1-regularized least-squares problem. Then, the candidate with the smallest projection error is taken as the tracking target. After that, tracking is continued using a Bayesian state inference framework. Two strategies are used to further improve the tracking performance. First, target templates are dynamically updated to capture appearance changes. Second, nonnegativity constraints are enforced to filter out clutter which negatively resembles tracking targets. We test the proposed approach on numerous sequences involving different types of challenges, including occlusion and variations in illumination, scale, and pose. The proposed approach demonstrates excellent performance in comparison with previously proposed trackers. We also extend the method for simultaneous tracking and recognition by introducing a static template set which stores target images from different classes. The recognition result at each frame is propagated to produce the final result for the whole video. The approach is validated on a vehicle tracking and classification task using outdoor infrared video sequences.

In this paper, we propose using sparse representation for recovering the illumination of a scene from a single image with cast shadows, given the geometry of the scene. The images with cast shadows can be quite complex and, therefore, cannot be well approximated by low-dimensional linear subspaces. However, it can be shown that the set of images produced by a Lambertian scene with cast shadows can be efficiently represented by a sparse set of images generated by directional light sources. We first model an image with cast shadows composed of a diffusive part (without cast shadows) and a residual part that captures cast shadows. Then, we express the problem in an l(1)-regularized least-squares formulation, with nonnegativity constraints (as light has to be non-negative at any point in space). This sparse representation enjoys an effective and fast solution thanks to recent advances in compressive sensing. In experiments on synthetic and real data, our approach performs favorably in comparison with several previously proposed methods.

In this paper we propose using bin-ratio information, which is collected from the ratios between bin values of histograms, for scene and category classification. To use such information, a new histogram dissimilarity, bin-ratio dissimilarity (BRD), is designed. We show that BRD provides several attractive advantages for category and scene classification tasks: First, BRD is robust to cluttering, partial occlusion and histogram normalization; Second, BRD captures rich co-occurrence information while enjoying a linear computational complexity; Third, BRD can be easily combined with other dissimilarity measures, such as L1and χ2, to gather complimentary information. We apply the proposed methods to category and scene classification tasks in the bag-of-words framework. The experiments are conducted on several widely tested datasets including PASCAL 2005, PASCAL 2008, Oxford flowers, and Scene-15 dataset. In all experiments, the proposed methods demonstrate excellent performance in comparison with previously reported solutions.

Based on multifractal analysis in wavelet pyramids of texture images, a new texture descriptor is proposed in this paper that implicitly combines information from both spatial and frequency domains. Beyond the traditional wavelet transform, a multi-oriented wavelet leader pyramid is used in our approach that robustly encodes the multi-scale information of texture edgels. Moreover, the resulting texture model shows empirically a strong power law relationship for nature textures, which can be characterized well by multifractal analysis. Combined with a statistics on affine invariant local patches, our proposed texture descriptor is robust to scale and rotation changes, more general geometrical transforms and illumination variations. In addition, the proposed texture descriptor is computationally efficient since it does not require many expensive processing steps, e.g., texton generation and cross-bin comparisons, which are often used by existing methods. As an application, the proposed descriptor is applied to texture classification and the experimental results on several public texture datasets verified the accuracy and efficiency of our descriptor.

Visual codebook has been popular in object classification as well as action analysis. However, its performance is often sensitive to the codebook size that is usually predefined. Moreover, the codebook generated by unsupervised methods, e.g., K-means, often suffers from the problem of ambiguity and weak efficiency. In other words, the visual codebook contains a lot of noisy and/or ambiguous words. In this paper, we propose a novel method to address these issues by constructing a compact but effective visual codebook using sparse reconstruction. Given a large codebook generated by K-means, we reformulate it in a sparse manner, and learn the weight of each word in the original visual codebook. Since the weights are sparse, they naturally introduce a new compact codebook. We apply this compact codebook to action recognition tasks and verify it on the widely used Weizmann action database. The experimental results show clearly the benefits of the proposed solution.

Conventional contour tracking algorithms with level set often use generative models to construct the energy function. For tracking through cluttered and noisy background, however, a generative model may not be discriminative enough. In this paper we integrate the discriminative methods into a level set framework when constructing the level set energy function. We train a set of weak classifiers to distinguish the object from the background. Each weak classifier is designed to select the most discriminative feature space and integrated via AdaBoost according to their training errors. We also introduce a novel interaction term to explore the correlation between pixels near the object edge. This term together with the discriminative model both enhance the discriminative power of the level set. The experimental results show that the contour tracked by our approach is more accurate than the conventional algorithms with the generative model. Our algorithm successfully tracks the object contour even in a cluttered environment.

We propose using multi-layer multiple instance learning (MMIL) for image set classification and applying it to the task of cannabis website classification. We treat each image as an instance in an image set, then each image is further viewed as containing instances of local image patches. This representation naturally extends traditional multiple instance learning (MIL) to multi-layers. We then show that, when using the set kernels for all layers, an MMIL problem can be flattened to a simple one-layer MIL. This flattening, when combined with quantized local image patch representation, drastically improves the computational efficiency by two orders. The flattened set kernel is further improved by weighted codewords and an exponential kernel. The proposed approach is applied to a cannabis website classification task, in which we collected a dataset containing more than 220,000 images from 600 websites. In the experiments our approach compares favorably with several state-of-the-art methods.

Face verification in the presence of age progression is an important problem that has not been widely addressed. In this paper, we study the problem by designing and evaluating discriminative approaches. These directly tackle verification tasks without explicit age modeling, which is a hard problem by itself. First, we find that the gradient orientation, after discarding magnitude information, provides a simple but effective representation for this problem. This representation is further improved when hierarchical information is used, which results in the use of the gradient orientation pyramid (GOP). When combined with a support vector machine GOP demonstrates excellent performance in all our experiments, in comparison with seven different approaches including two commercial systems. Our experiments are conducted on the FGnet dataset and two large passport datasets, one of them being the largest ever reported for recognition tasks. Second, taking advantage of these datasets, we empirically study how age gaps and related issues (including image quality, spectacles, and facial hair) affect recognition algorithms. We found surprisingly that the added difficulty of verification produced by age gaps becomes saturated after the gap is larger than four years, for gaps of up to ten years. In addition, we find that image quality and eyewear present more of a challenge than facial hair.

Recently, we proposed marginal space learning (MSL) as a generic approach for automatic detection of 3D anatomical structures in many medical imaging modalities. To accurately localize a 3D object, we need to estimate nine parameters (three for position, three for orientation, and three for anisotropic scaling). Instead of uniformly searching the original nine-dimensional parameter space, only low-dimensional marginal spaces are uniformly searched in MSL, which significantly improves the speed. In many real applications, a strong correlation may exist among parameters in the same marginal spaces. For example, a large object may have large scaling values along all directions. In this paper, we propose constrained MSL to exploit this correlation for further speed-up. As another major contribution, we propose to use quaternions for 3D orientation representation and distance measurement to overcome the inherent drawbacks of Euler angles in the original MSL. The proposed method has been tested on three 3D anatomical structure detection problems in medical images, including liver detection in computed tomography (CT) volumes, and left ventricle detection in both CT and ultrasound volumes. Experiments on the largest datasets ever reported show that constrained MSL can improve the detection speed up to 14 times, while achieving comparable or better detection accuracy. It takes less than half a second to detect a 3D anatomical structure in a volume.

We propose a novel framework for contour based object detection and recognition, which we formulate as a joint contour fragment grouping and labeling problem. For a given set of contours of model shapes, we simultaneously perform selection of relevant contour fragments in edge images, grouping of the selected contour fragments, and their matching to the model contours. The inference in all these steps is performed using particle filters (PF) but with static observations. Our approach needs one example shape per class as training data. The PF framework combined with decomposition of model contour fragments to part bundles allows us to implement an intuitive search strategy for the target contour in a clutter of edge fragments. First a rough sketch of the model shape is identified, followed by fine tuning of shape details. We show that this framework yields not only accurate object detections but also localizations in real cluttered images.

In this paper we propose a robust visual tracking method by casting tracking as a sparse approximation problem in a particle filter framework. In this framework, occlusion, corruption and other challenging issues are addressed seamlessly through a set of trivial templates. Specifically, to find the tracking target at a new frame, each target candidate is sparsely represented in the space spanned by target templates and trivial templates. The sparsity is achieved by solving an ℓ1-regularized least squares problem. Then the candidate with the smallest projection error is taken as the tracking target. After that, tracking is continued using a Bayesian state inference framework in which a particle filter is used for propagating sample distributions over time. Two additional components further improve the robustness of our approach: 1) the nonnegativity constraints that help filter out clutter that is similar to tracked targets in reversed intensity patterns, and 2) a dynamic template update scheme that keeps track of the most representative templates throughout the tracking procedure. We test the proposed approach on five challenging sequences involving heavy occlusions, drastic illumination changes, and large pose variations. The proposed approach shows excellent performance in comparison with three previously proposed trackers.

Scenes with cast shadows can produce complex sets of images. These images cannot be well approximated by low-dimensional linear subspaces. However, in this paper we show that the set of images produced by a Lambertian scene with cast shadows can be efficiently represented by a sparse set of images generated by directional light sources. We first model an image with cast shadows as composed of a diffusive part (without cast shadows) and a residual part that captures cast shadows. Then, we express the problem in an ℓ1-regularized least squares formulation, with nonnegativity constraints. This sparse representation enjoys an effective and fast solution, thanks to recent advances in compressive sensing. In experiments on both synthetic and real data, our approach performs favorably in comparison to several previously proposed methods.

Predicting the age of a person through face image analysis holds the potential to drive an extensive array of real world applications from human computer interaction and security to advertising and multimedia. In this paper the first application of the random forest for age regression is proposed. This method offers the advantage of few parameters that are relatively easy to initialize. Our method learns salient anthropometric quantities without a prior model. Significant implications include a dramatic reduction in training time while maintaining high regression accuracy throughout human development.

Thumbnail cropping helps improve thumbnail readability by cropping images before shrinking them. In this paper we propose a learning based method for automatic thumbnail cropping. To this end, we use a support vector machine to learn a discriminative model that simultaneously captures the saliency distribution and spatial priors. The model is then used to determine the best cropping rectangle. The proposed approach improves traditional saliency based cropping techniques by introducing the spatial priors, which is automatically learned through learning process. The new method is tested on images from the PASCAL08 dataset, where it outperforms previous saliency based cropping.

In this paper we present a hierarchical, learning-based approach for automatic and accurate liver segmentation from 3D CT volumes. We target CT volumes that come from largely diverse sources (e.g., diseased in six different organs) and are generated by different scanning protocols (e.g., contrast and non-contrast, various resolution and position). Three key ingredients are combined to solve the segmentation problem. First, a hierarchical framework is used to efficiently and effectively monitor the accuracy propagation in a coarse-to-fine fashion. Second, two new learning techniques, marginal space learning and steerable features, are applied for robust boundary inference. This enables handling of highly heterogeneous texture pattern. Third, a novel shape space initialization is proposed to improve traditional methods that are limited to similarity transformation. The proposed approach is tested on a challenging dataset containing 174 volumes. Our approach not only produces excellent segmentation accuracy, but also runs about fifty times faster than state-of-the-art solutions [7, 9].

We propose using the proximity distribution of vector- quantized local feature descriptors for object and category recognition. To this end, we introduce a novel "proximity distribution kernel" that naturally combines local geometric as well as photometric information from images. It satisfies Mercer's condition and can therefore be readily combined with a support vector machine to perform visual categorization in a way that is insensitive to photometric and geometric variations, while retaining significant discriminative power. In particular, it improves on the results obtained both with geometrically unconstrained "bags of features" approaches, as well as with over-constrained "affine procrustes." Indeed, we test this approach on several challenging data sets, including Graz-01, Graz-02, and the PASCAL challenge. We registered the average performance at 91.5% on Graz-01, 82.7% on Graz-02, and 74.5% on PASCAL. Our approach is designed to enforce and exploit geometric consistency among objects in the same category; therefore, it does not improve the performance of existing algorithms on datasets where the data is already roughly aligned and scaled. Our method has the potential to be extended to more complex geometric relationships among local features, as we illustrate in the experiments.

In this paper we study face recognition across ages within a real passport photo verification task. First, we propose using the gradient orientation pyramid for this task. Discarding the gradient magnitude and utilizing hierarchical techniques, we found that the new descriptor yields a robust and discriminative representation. With the proposed descriptor, we model face verification as a two-class problem and use a support vector machine as a classifier. The approach is applied to two passport data sets containing more than 1,800 image pairs from each person with large age differences. Although simple, our approach outperforms previously tested Bayesian technique and other descriptors, including the intensity difference and gradient with magnitude. In addition, it works as well as two commercial systems. Second, for the first time, we empirically study how age differences affect recognition performance. Our experiments show that, although the aging process adds difficulty to the recognition task, it does not surpass illumination or expression as a confounding factor.

Part structure and articulation are of fundamental importance in computer and human vision. We propose using the inner-distance to build shape descriptors that are robust to articulation and capture part structure. The inner-distance is defined as the length of the shortest path between landmark points within the shape silhouette. We show that it is articulation insensitive and more effective at capturing part structures than the Euclidean distance. This suggests that the inner-distance can be used as a replacement for the Euclidean distance to build more accurate descriptors for complex shapes, especially for those with articulated parts. In addition, texture information along the shortest path can be used to further improve shape classification. With this idea, we propose three approaches to using the inner-distance. The first method combines the inner-distance and multidimensional scaling (MDS) to build articulation invariant signatures for articulated shapes. The second method uses the inner-distance to build a new shape descriptor based on shape contexts. The third one extends the second one by considering the texture information along shortest paths. The proposed approaches have been tested on a variety of shape databases, including an articulated shape data set, MPEG7 CE-Shape-1, Kimia silhouettes, the ETH-80 data set, two leaf data sets, and a human motion silhouette data set. In all the experiments, our methods demonstrate effective performance compared with other algorithms.

We propose EMD-L1: a fast and exact algorithm for computing the Earth Mover's Distance (EMD) between a pair of histograms. The efficiency of the new algorithm enables its application to problems that were previously prohibitive due to high time complexities. The proposed EMD-L1 significantly simplifies the original linear programming formulation of EMD. Exploiting the L1 metric structure, the number of unknown variables in EMD-L1 is reduced to O(N) from O(N2) of the original EMD for a histogram with N bins. In addition, the number of constraints is reduced by half and the objective function of the linear program is simplified. Formally, without any approximation, we prove that the EMD-L1 formulation is equivalent to the original EMD with a L1 ground distance. To perform the EMD-L1 computation, we propose an efficient tree-based algorithm, Tree-EMD. Tree-EMD exploits the fact that a basic feasible solution of the simplex algorithm-based solver forms a spanning tree when we interpret EMD-L1 as a network flow optimization problem. We empirically show that this new algorithm has an average time complexity of O(N2), which significantly improves the best reported supercubic complexity of the original EMD. The accuracy of the proposed methods is evaluated by experiments for two computation-intensive problems: shape recognition and interest point matching using multidimensional histogram-based local features. For shape recognition, EMD-L1 is applied to compare shape contexts on the widely tested MPEG7 shape data set, as well as an articulated shape data set. For interest point matching, SIFT, shape context and spin image are tested on both synthetic and real image pairs with large geometrical deformation, illumination change, and heavy intensity noise. The results demonstrate that our EMD-L1-based solutions outperform previously reported state-of-the-art features and distance measures in solving the two tasks.

In this paper we propose diffusion distance, a new dissimilarity measure between histogram-based descriptors. We define the difference between two histograms to be a temperature field. We then study the relationship between histogram similarity and a diffusion process, showing how diffusion handles deformation as well as quantization effects. As a result, the diffusion distance is derived as the sum of dissimilarities over scales. Being a cross-bin histogram distance, the diffusion distance is robust to deformation, lighting change and noise in histogram-based local descriptors. In addition, it enjoys linear computational complexity which significantly improves previously proposed cross-bin distances with quadratic complexity or higher. We tested the proposed approach on both shape recognition and interest point matching tasks using several multi-dimensional histogram-based descriptors including shape context, SIFT, and spin images. In all experiments, the diffusion distance performs excellently in both accuracy and efficiency in comparison with other state-of-the-art distance measures. In particular, it performs as accurately as the Earth Mover’s Distance with much greater efficiency.

We propose using the inner-distance between landmark points to build shape descriptors. The inner-distance is defined as the length of the shortest path between landmark points within the shape silhouette. We show that the inner-distance is articulation insensitive and more effective at capturing complex shapes with part structures than Euclidean distance. To demonstrate this idea, it is used to build a new shape descriptor based on shape contexts. After that, we design a dynamic programming based method for shape matching and comparison. We have tested our approach on a variety of shape databases including an articulated shape dataset, MPEG7 CE-Shape-1, Kimia silhouettes, a Swedish leaf database and a human motion silhouette dataset. In all the experiments, our method demonstrates effective performance compared with other algorithms.

We propose a novel framework to build descriptors of local intensity that are invariant to general deformations. In this framework, an image is embedded as a 2D surface in 3D space, with intensity weighted relative to distance in x-y. We show that as this weight increases, geodesic distances on the embedded surface are less affected by image deformations. In the limit, distances are deformation invariant. We use geodesic sampling to get neighborhood samples for interest points, and then use a geodesic-intensity histogram (GIH) as a deformation invariant local descriptor. In addition to its invariance, the new descriptor automatically finds its support region. This means it can safely gather information from a large neighborhood to improve discriminability. Furthermore, we propose a matching method for this descriptor that is invariant to affine lighting changes. We have tested this new descriptor on interest point matching for two data sets, one with synthetic deformation and lighting change, and another with real non-affine deformations. Our method shows promising matching results compared to several other approaches

Thumbnail images provide users of image retrieval and browsing systems with a method for quickly scanning large numbers of images. Recognizing the objects in an image is important in many retrieval tasks, but thumbnails generated by shrinking the original image often render objects illegible. We study the ability of computer vision systems to detect key components of images so that automated cropping, prior to shrinking, can render objects more recognizable. We evaluate automatic cropping techniques 1) based on a general method that detects salient portions of images, and 2) based on automatic face detection. Our user study shows that these methods result in small thumbnails that are substantially more recognizable and easier to find in the context of visual search.