Little research explores the correlation between the expressive ability and generalization ability of the low-rank adaptation (LoRA). Sharpness-Aware Minimization (SAM) improves model generalization for both Convolutional Neural Networks (CNNs) and Transformers by encouraging convergence to locally flat minima. However, the connection between sharpness and generalization has not been fully explored for LoRA due to the lack of tools to either empirically seek flat minima or develop theoretical methods. In this work, we propose Flat Minima LoRA (FMLoRA) and its efficient version i.e., EFMLoRA, to seek flat minima for LoRA. Concretely, we theoretically demonstrate that perturbations in the full parameter space can be transferred to the low-rank subspace. This approach eliminates the potential interference introduced by perturbations across multiple matrices in the low-rank subspace. Our extensive experiments on large language models and vision-language models demonstrate that EFMLoRA achieves optimization efficiency comparable to that of LoRA while simultaneously attaining comparable or even better performance. For example, on the GLUE dataset with RoBERTa-large, EFMLoRA outperforms LoRA and full fine-tuning by 1.0% and 0.5% on average, respectively. On vision-language models e.g., Qwen-VL-Chat, there are performance improvements of 1.5% and 1.0% on the SQA and VizWiz datasets, respectively. These empirical results also verify that the generalization of LoRA is closely related to sharpness, which is omitted by previous methods.

Industrial gas leakage detection is critically important for safety and environmental protection. While infrared imaging enables detection of invisible gases, two challenges remain: existing datasets lack realistic industrial scenarios, and current methods struggle to distinguish gas plumes from background interferences or segment discontinuous gas distributions. This paper introduces a benchmark comprising an Industrial RGB-Thermal Dataset (IRTD) with gas emission and leakage data from laboratory and industrial sites. A VLM-assisted RGBThermal detection framework with a Cross-Attention based Feature Difference (CAFD) module is designed to enhance gasspecific feature differentiation by computing inter-modal feature discrepancies. Evaluations on public datasets and IRTD demonstrate state-of-the-art results.

In recent times, geometric foundation models have demonstrated remarkable performance in depth estimation tasks, benefiting from exposure to large-scale data that enables the learning of intricate geometric structures and spatial dependencies. However, their large parameter sizes and high computational complexity pose significant challenges in meeting the efficiency requirements of downstream surgical applications. Consequently, the design of a high-performance yet lightweight monocular depth estimator has become a focal point of research. To this end, we harness the rich geometric priors encoded in geometric foundation models and introduce a novel trinity distillation scheme that transfers geometric knowledge across three complementary dimensions, namely spatial, spectral and gradient, into a compact depth estimator. To further enhance prediction quality, we develop a semantic distribution alignment strategy to effectively suppress pseudo-texture artifacts arising from the limited semantic representation capability of the lightweight estimator. Extensive experiments on the SCARED, SERV-CT, Hamlyn, and C3VD datasets demonstrate that the proposed method either surpasses or achieves comparable performance to previous state-of-the-art competitors, with a smaller model size and reduced computational overhead. Code will be available at: https://github.com/ShuweiShao/LiteNet.

Sharpness-Aware Minimization (SAM) has emerged as a promising approach for effectively reducing the generalization error. However, SAM incurs twice the computational cost compared to the base optimizer (e.g., SGD). We propose Asymptotic Unbiased data sampling to accelerate SAM (AUSAM), which maintains the model's generalization capacity while significantly enhancing computational efficiency. Concretely, we probabilistically sample a subset of data points beneficial for SAM optimization based on a theoretically guaranteed criterion, i.e., the Gradient Norm of each Sample (GNS). We further approximate the GNS by evaluating the difference in loss values before and after perturbation in SAM. As a plug-and-play, architecture-agnostic method, our approach consistently accelerates SAM across various tasks and networks, i.e., classification, human pose estimation, and network quantization. On CIFAR-10/100 and Tiny-ImageNet, AUSAM achieves results comparable to SAM while providing a speedup of over 70%. By adjusting hyperparameters, AUSAM can match the speed of the base optimizer while significantly surpassing the base optimizer's performance. Compared to recent dynamic data pruning methods, AUSAM is better suited for SAM and excels in maintaining performance. Additionally, AUSAM accelerates optimization in human pose estimation and model quantization without sacrificing performance, demonstrating its broad practicality.

Convolutional neural networks (CNNs) have been playing a dominant role in computer vision. However, the existing approaches of using local window modeling in popular CNNs lack flexibility and hinder their ability to capture long-range dependencies of objects in an image. To overcome these limitations, we propose a novel CNN architecture, termed Dynamic Clustering Convolutional Neural Network (DCCNeXt). The proposed DCCNeXt takes a unique approach by employing global clustering to group image patches with similar semantics into clusters that are then convolved using the shared convolution kernels. To address the high computational complexity of global clustering, the feature vectors from each patch's subspace are extracted for efficient clustering, which makes the proposed model widely compatible with the downstream vision tasks. The extensive experiments of image classification, object detection, instance segmentation, and semantic segmentation on the benchmark datasets demonstrate that the proposed DCCNeXt outperforms the mainstream Convolutional Neural Networks (CNNs), Vision Transformers (ViTs), Vision Multi-layer Perceptrons (MLPs), Vision Graph Neural Networks (GNNs), and Vision Mambas. We anticipate that this study will provide a new perspective and a promising avenue for the design of convolutional neural networks.

Graph Neural Networks (GNNs) have recently gained widespread attention as a successful tool for analyzing graph-structured data. However, imperfect graph structure with noisy links lacks enough robustness and may damage graph representations, therefore limiting the GNNs' performance in practical tasks. Moreover, existing generative architectures fail to fit discriminative graph-related tasks. To tackle these issues, we introduce an unsupervised method based on a joint of generative training and discriminative training to learn graph structure and representation, aiming to improve the discriminative performance of generative models. We propose an Energy-based Contrastive Learning (ECL) guided Graph Structure Refinement (GSR) framework, denoted as ECL-GSR. To our knowledge, this is the first work to combine energy-based models with contrastive learning for GSR. Specifically, we leverage ECL to approximate the joint distribution of sample pairs, which increases the similarity between representations of positive pairs while reducing the similarity between negative ones. Refined structure is produced by augmenting and removing edges according to the similarity metrics among node representations. Extensive experiments demonstrate that ECL-GSR outperforms the state-of-the-art on eight benchmark datasets in node classification. ECL-GSR achieves faster training with fewer samples and memories against the leading baseline, highlighting its simplicity and efficiency in downstream tasks.

Deep learning has significantly advanced the object detection field. However, tiny object detection (TOD) remains a challenging problem. We provide a new analysis method to examine the TOD challenge through occlusion-based attribution analysis in the frequency domain. We observe that tiny objects become less distinct after feature encoding and can benefit from the removal of high-frequency information. In this paper, we propose a novel approach named Spectral Enhancement for Tiny object detection (SET), which amplifies the frequency signatures of tiny objects in a heterogeneous architecture. SET includes two modules. The Hierarchical Background Smoothing (HBS) module suppresses high-frequency noise in the background through adaptive smoothing operations. The Adversarial Perturbation Injection (API) module leverages adversarial perturbations to increase feature saliency in critical regions and prompt the refinement of object features during training. Extensive experiments on four datasets demonstrate the effectiveness of our method. Especially, SET boosts the prior art RFLA by 3.2% AP on the AI-TOD dataset.

Feature matching methods for unsupervised anomaly detection have demonstrated impressive performance. Existing methods primarily rely on self-supervised training and handcrafted matching schemes for task adaptation. However, they can only achieve an inferior feature representation for anomaly detection because the feature extraction and matching modules are separately trained. To address these issues, we propose a Differentiable Feature Matching (DFM) framework for joint optimization of the feature extractor and the matching head. DFM transforms nearest-neighbor matching into a pooling-based module and embeds it within a Feature Matching Network (FMN). This design enables end-to-end feature extraction and feature matching module training, thus providing better feature representation for anomaly detection tasks. DFM is generic and can be incorporated into existing feature-matching methods. We implement DFM with various backbones and conduct extensive experiments across various tasks and datasets, demonstrating its effectiveness. Notably, we achieve state-of-the-art results in the continual anomaly detection task with instance-AUROC improvement of up to 3.9% and pixel-AP improvement of up to 5.5%.

Leveraging the complementary characteristics of visible (RGB) and infrared (IR) imagery offers significant potential for improving object detection. In this paper, we propose WaveMamba, a cross-modality fusion method that efficiently integrates the unique and complementary frequency features of RGB and IR decomposed by Discrete Wavelet Transform (DWT). An improved detection head incorporating the Inverse Discrete Wavelet Transform (IDWT) is also proposed to reduce information loss and produce the final detection results. The core of our approach is the introduction of WaveMamba Fusion Block (WMFB), which facilitates comprehensive fusion across low-/high-frequency sub-bands. Within WMFB, the Low-frequency Mamba Fusion Block (LMFB), built upon the Mamba framework, first performs initial low-frequency feature fusion with channel swapping, followed by deep fusion with an advanced gated attention mechanism for enhanced integration. High-frequency features are enhanced using a strategy that applies an ``absolute maximum" fusion approach. These advancements lead to significant performance gains, with our method surpassing state-of-the-art approaches and achieving average mAP improvements of 4.5% on four benchmarks.

Zero-shot medical detection can further improve detection performance without relying on annotated medical images even upon the fine-tuned model, showing great clinical value. Recent studies leverage grounded vision-language models (GLIP) to achieve this by using detailed disease descriptions as prompts for the target disease name during the inference phase. However, these methods typically treat prompts as equivalent context to the target name, making it difficult to assign specific disease knowledge based on visual information, leading to a coarse alignment between images and target descriptions. In this paper, we propose StructuralGLIP, which introduces an auxiliary branch to encode prompts into a latent knowledge bank layer-by-layer, enabling more context-aware and fine-grained alignment. Specifically, in each layer, we select highly similar features from both the image representation and the knowledge bank, forming structural representations that capture nuanced relationships between image patches and target descriptions. These features are then fused across modalities to further enhance detection performance. Extensive experiments demonstrate that StructuralGLIP achieves a +4.1\% AP improvement over prior state-of-the-art methods across seven zero-shot medical detection benchmarks, and consistently improves fine-tuned models by +3.2\% AP on endoscopy image datasets.

Co-training can achieve parameter-efficient multi-task models but remains unexplored for quantization-aware training. Our investigation shows that directly introducing co-training into existing quantization-aware training (QAT) methods results in significant performance degradation. Our experimental study identifies that the primary issue with existing QAT methods stems from the inadequate activation quantization scales for the co-training framework. To address this issue, we propose Task-Specific Scales Quantization for Multi-Task Co-Training (TSQ-MTC) to tackle mismatched quantization scales. Specifically, a task-specific learnable multi-scale activation quantizer (TLMAQ) is incorporated to enrich the representational ability of shared features for different tasks. Additionally, we find that in the deeper layers of the Transformer model, the quantized network suffers from information distortion within the attention quantizer. A structure-based layer-by-layer distillation (SLLD) is then introduced to ensure that the quantized features effectively preserve the information from their full-precision counterparts. Our extensive experiments in two co-training scenarios demonstrate the effectiveness and versatility of TSQ-MTC. In particular, we successfully achieve a 4-bit quantized low-level visual foundation model based on IPT, which attains a PSNR comparable to the full-precision model while offering a $7.99\times$ compression ratio in the $\times4$ super-resolution task on the Set5 benchmark.

Recent advances in 3D generation have been remarkable, with methods such as DreamFusion leveraging large-scale text-to-image diffusion-based models to guide 3D object generation. These methods enable the synthesis of detailed and photorealistic textured objects. However, the appearance of 3D objects produced by such text-to-3D models is often unpredictable, and it is hard for single-image-to-3D methods to deal with images lacking a clear subject, complicating the generation of appearance-controllable 3D objects from complex images. To address these challenges, we present IPDreamer, a novel method that captures intricate appearance features from complex **I**mage **P**rompts and aligns the synthesized 3D object with these extracted features, enabling high-fidelity, appearance-controllable 3D object generation. Our experiments demonstrate that IPDreamer consistently generates high-quality 3D objects that align with both the textual and complex image prompts, highlighting its promising capability in appearance-controlled, complex 3D object generation.

State space models (SSMs) have recently garnered significant attention in computer vision. However, due to the unique characteristics of image data, adapting SSMs from natural language processing to computer vision has not outperformed the state-of-the-art convolutional neural networks (CNNs) and Vision Transformers (ViTs). Existing vision SSMs primarily leverage manually designed scans to flatten image patches into sequences locally or globally. This approach disrupts the original semantic spatial adjacency of the image and lacks flexibility, making it difficult to capture complex image structures. To address this limitation, we propose Dynamic Adaptive Scan (DAS), a data-driven method that adaptively allocates scanning orders and regions. This enables more flexible modeling capabilities while maintaining linear computational complexity and global modeling capacity. Based on DAS, we further propose the vision backbone DAMamba, which significantly outperforms popular vision Mamba models in vision tasks such as image classification, object detection, instance segmentation, and semantic segmentation. Notably, it surpasses some of the latest state-of-the-art CNNs and ViTs.

In real-world scenarios, the number of training samples across classes usually subjects to a long-tailed distribution. The conventionally trained network may achieve unexpected inferior performance on the rare class compared to the frequent class. Most previous works attempt to rectify the network bias from the data-level or from the classifier-level. Differently, in this paper, we identify that the bias towards the frequent class may be encoded into features, i.e., the rare-specific features which play a key role in discriminating the rare class are much weaker than the frequent-specific features. Based on such an observation, we introduce a simple yet effective approach, normalizing the parameters of Batch Normalization (BN) layer to explicitly rectify the feature bias. To achieve this end, we represent theWeight/Bias parameters of a BN layer as a vector, normalize it into a unit one and multiply the unit vector by a scalar learnable parameter. Through decoupling the direction and magnitude of parameters in BN layer to learn, the Weight/Bias exhibits a more balanced distribution and thus the strength of features becomes more even. Extensive experiments on various long-tailed recognition benchmarks (i.e., CIFAR-10/100-LT, ImageNet-LT and iNaturalist 2018) show that our method outperforms previous state-of-the-arts remarkably.

Cross-modality object detection aims to fuse complementary information from different modalities to improve model performance, which achieves a wider range of applications. However, traditional cross-modality fusion methods, based on CNN or Transformer, inadequately address the issue of pseudo-target information, which causes model attention dispersion to degrade object detection performance. In this paper, we investigate a novel cross-modality fusion approach by associating cross-modal features in a hidden state space based on an improved Mamba with a gating attention mechanism. We propose theFusion-Mamba Block(FMB), designed to map cross-modal features into a hidden state space for interaction, thereby refining the model’s attention on true target areas and enhancing overall performance. The FMB comprises two key modules: State Space Channel Swapping (SSCS) module, which facilitates the fusion of shallow features, and Dual State Space Fusion (DSSF) module, which enables deep fusion and effectively suppresses pseudo-target information within the hidden state space. Our proposed method outperforms state-of-the-art approaches, achieving improvements of 5.9%, 3.5% and 2.1% mAP on$M^{3}$FD, DroneVehicle and FLIR-Aligned, respectively. To the best of our knowledge, this work establishes a new baseline for cross-modality object detection, providing a robust foundation for future research in this area.

Fine-grained visual categorization (FGVC) relies on hierarchical features extracted by deep convolutional neural networks (CNNs) to recognize closely alike objects. Particularly, shallow layer features containing rich spatial details are vital for specifying subtle differences between objects but are usually inadequately optimized due to gradient vanishing during backpropagation. In this article, hierarchical self-distillation (HSD) is introduced to generate well-optimized CNNs features for accurate fine-grained categorization. HSD inherits from the widely applied deep supervision and implements multiple intermediate losses for reinforced gradients. Besides that, we observe that the hard (one-hot) labels adopted for intermediate supervision hurt the performance of FGVC by enforcing overstrict supervision. As a solution, HSD seeks self-distillation where soft predictions generated by deeper layers of the network are hierarchically exploited to supervise shallow parts. Moreover, self-information entropy loss (SIELoss) is designed in HSD to adaptively soften intermediate predictions and facilitate better convergence. In addition, the gradient detached fusion (GDF) module is incorporated to produce an ensemble result with multiscale features via effective feature fusion. Extensive experiments on four challenging fine-grained datasets show that, with neglectable parameter increase, the proposed HSD framework and the GDF module both bring significant performance gains over different backbones, which also achieves state-of-the-art classification performance.

Most of the existing fusion algorithms are not robust to unregistered input images. Even after image registration, nonlinear nonregistration may persist in the local areas of the images, leading to poor quality in the fused image. So, as to tackle these challenges, a progressive remote sensing image registration and fusion network is proposed in this article, and named PRF-Net, which is particularly useful when two images are from different platforms. First, a registration network is designed to register the input image patches, which includes a global spatial transform network (GSTN) and a local spatial warp network (LSWN). The GSTN is primarily used for coarse registration, applying rigid transformation to globally align the input images. After coarse registration, the preliminarily registered moving image is input into the LSWN for local fine-tuning to maximize correlation between the input image patches. Subsequently, the fine registered images are degraded and input into the fusion network to generate the fused image. To maintain sufficient spectral and spatial information of the fused image, a multiscale feature extraction (MSFE) block with a highly interpretable spatial details attention (SDA) block is designed, which can enhance the ability of fusion network to extract and preserve spatial details and spectral information. Three groups of experiments conducted on four types of remote sensing images give evidence of that the proposed PRF-Net exhibits excellent performance in both reduced and full resolutions, showcasing its outstanding registration and fusion quality.

While the deep convolutional neural network (DCNN) has achieved overwhelming success in various vision tasks, its heavy computational and storage overhead hinders the practical use of resource-constrained devices. Recently, compressing DCNN models has attracted increasing attention, where binarization-based schemes have generated great research popularity due to their high compression rate. In this article, we propose modulated convolutional networks (MCNs) to obtain binarized DCNNs with high performance. We lead a new architecture in MCNs to efficiently fuse the multiple features and achieve a similar performance as the full-precision model. The calculation of MCNs is theoretically reformulated as a discrete optimization problem to build binarized DCNNs, for the first time, which jointly consider the filter loss, center loss, and softmax loss in a unified framework. Our MCNs are generic and can decompose full-precision filters in DCNNs, e.g., conventional DCNNs, VGG, AlexNet, ResNets, or Wide-ResNets, into a compact set of binarized filters which are optimized based on a projection function and a new updated rule during the backpropagation. Moreover, we propose modulation filters (M-Filters) to recover filters from binarized ones, which lead to a specific architecture to calculate the network model. Our proposed MCNs substantially reduce the storage cost of convolutional filters by a factor of 32 with a comparable performance to the full-precision counterparts, achieving much better performance than other state-of-the-art binarized models.

Consistent editing of real images is a challenging task, as it requires performing non-rigid edits (e.g., changing postures) to the main objects in the input image without changing their identity or attributes. To guarantee consistent attributes, some existing methods fine-tune the entire model or the textual embedding for structural consistency, but they are time-consuming and fail to perform non-rigid edits. Other works are tuning-free, but their performances are weakened by the quality of Denoising Diffusion Implicit Model (DDIM) reconstruction, which often fails in real-world scenarios. In this paper, we present a novel approach called Tuning-free Inversion-enhanced Control (TIC), which directly correlates features from the inversion process with those from the sampling process to mitigate the inconsistency in DDIM reconstruction. Specifically, our method effectively obtains inversion features from the key and value features in the self-attention layers, and enhances the sampling process by these inversion features, thus achieving accurate reconstruction and content-consistent editing. To extend the applicability of our method to general editing scenarios, we also propose a mask-guided attention concatenation strategy that combines contents from both the inversion and the naive DDIM editing processes. Experiments show that the proposed method outperforms previous works in reconstruction and consistent editing, and produces impressive results in various settings.

Federated learning (FL) is a machine learning paradigm in which distributed local nodes collaboratively train a central model without sharing individually held private data. Existing FL methods either iteratively share local model parameters or deploy co-distillation. However, the former is highly susceptible to private data leakage, and the latter design relies on the prerequisites of task-relevant real data. Instead, we propose a data-free FL framework based on local-to-central collaborative distillation with direct input and output space exploitation. Our design eliminates any requirement of recursive local parameter exchange or auxiliary task-relevant data to transfer knowledge, thereby giving direct privacy control to local users. In particular, to cope with the inherent data heterogeneity across locals, our technique learns to distill input on which each local model produces consensual yet unique results to represent each expertise. Our proposed FL framework achieves notable privacy-utility trade-offs with extensive experiments on image classification and segmentation tasks under various real-world heterogeneous federated learning settings on both natural and medical images. Code is available at https://github.com/lsl001006/FedIOD.

Vision transformers (ViTs) quantization offers a promising prospect to facilitate deploying large pre-trained networks on resource-limited devices. Fully-binarized ViTs (Bi-ViT) that pushes the quantization of ViTs to its limit remain largely unexplored and a very challenging task yet, due to their unacceptable performance. Through extensive empirical analyses, we identify the severe drop in ViT binarization is caused by attention distortion in self-attention, which technically stems from the gradient vanishing and ranking disorder. To address these issues, we first introduce a learnable scaling factor to reactivate the vanished gradients and illustrate its effectiveness through theoretical and experimental analyses. We then propose a ranking-aware distillation method to rectify the disordered ranking in a teacher-student framework. Bi-ViT achieves significant improvements over popular DeiT and Swin backbones in terms of Top-1 accuracy and FLOPs. For example, with DeiT-Tiny and Swin-Tiny, our method significantly outperforms baselines by 22.1% and 21.4% respectively, while 61.5x and 56.1x theoretical acceleration in terms of FLOPs compared with real-valued counterparts on ImageNet. Our codes and models are attached on https://github.com/YanjingLi0202/Bi-ViT/ .

DEtection TRansformer (DETR)-based models have achieved remarkable performance. However, they are accompanied by a large computation overhead cost, which significantly prevents their applications on resource-limited devices. Prior arts attempt to reduce the computational burden of DETR using low-bit quantization, while these methods sacrifice a severe significant performance on weight-activation-attention low-bit quantization. We observe that the number of matching queries and positive samples affect much on the representation capacity of queries in DETR, while quantifying queries of DETR further reduces its representational capacity, thus leading to a severe performance drop. We introduce a new quantization strategy based on Auxiliary Queries for DETR (AQ-DETR), aiming to enhance the capacity of quantized queries. In addition, a layer-by-layer distillation is proposed to reduce the quantization error between quantized attention and full-precision counterpart. Through our extensive experiments on large-scale open datasets, the performance of the 4-bit quantization of DETR and Deformable DETR models is comparable to full-precision counterparts.

Brain signal visualization has emerged as an active research area, serving as a critical interface between the human visual system and computer vision models. Diffusion-based methods have recently shown promise in analyzing functional magnetic resonance imaging (fMRI) data, including the reconstruction of high-quality images consistent with original visual stimuli. Nonetheless, it remains a critical challenge to effectively harness the semantic and silhouette information extracted from brain signals. In this paper, we propose a novel approach, termed as Controllable Mind Visual Diffusion Model (CMVDM). Specifically, CMVDM first extracts semantic and silhouette information from fMRI data using attribute alignment and assistant networks. Then, a control model is introduced in conjunction with a residual block to fully exploit the extracted information for image synthesis, generating high-quality images that closely resemble the original visual stimuli in both semantic content and silhouette characteristics. Through extensive experimentation, we demonstrate that CMVDM outperforms existing state-of-the-art methods both qualitatively and quantitatively. Our code is available at https://github.com/zengbohan0217/CMVDM.

The remarkable performance of Vision Transformers (ViTs) typically requires an extremely large training cost. Existing methods have attempted to accelerate the training of ViTs, yet typically disregard method universality with accuracy dropping. Meanwhile, they break the training consistency of the original transformers, including the consistency of hyperparameters, architecture, and strategy, which prevents them from being widely applied to different Transformer networks. In this paper, we propose a novel token growth scheme Token Expansion (termed ToE) to achieve consistent training acceleration for ViTs. We introduce an “initialization-expansion-merging” pipeline to maintain the integrity of the intermediate feature distribution of original transformers, preventing the loss of crucial learnable information in the training process. ToE can not only be seamlessly integrated into the training and fine-tuning process of transformers (e.g., DeiT and LV-ViT), but also effective for efficient training frameworks (e.g., EfficientTrain), without twisting the original training hyperparameters, architecture, and introducing additional training strategies. Extensive experiments demonstrate that ToE achieves about 1.3× faster for the training of ViTs in a lossless manner, or even with performance gains over the full-token training baselines. Code is available at https://github.com/Osilly/TokenExpansion.

3D face reconstruction aims at generating high-fidelity 3D face shapes and textures from single-view or multi-view images. However, current prevailing facial texture generation methods generally suffer from low-quality texture, identity information loss, and inadequate handling of occlusions. To solve these problems, we introduce an Identity-Conditioned Latent Diffusion Model for face UV-texture generation (UV-IDM) to generate photo-realistic textures based on the Basel Face Model (BFM). UV-IDM leverages the powerful texture generation capacity of a latent diffusion model (LDM) to obtain detailed facial textures. To preserve the identity during the reconstruction procedure, we design an identity-conditioned module that can utilize any in-the-wild image as a robust condition for the LDM to guide texture generation. UV-IDM can be easily adapted to different BFM-based methods as a high-fidelity texture generator. Furthermore, in light of the limited accessibility of most existing UV-texture datasets, we build a large-scale and publicly available UV-texture dataset based on BFM, termed BFM-UV. Extensive experiments show that our UV-IDM can generate high-fidelity textures in 3D face reconstruction within seconds while maintaining image consistency, bringing new state-of-the-art performance in facial texture generation.

Recent advances in vision-language models like Stable Diffusion have shown remarkable power in creative image synthesis and editing. However, most existing text-to-image editing methods encounter two obstacles: First, the text prompt needs to be carefully crafted to achieve good results, which is not intuitive or user-friendly. Second, they are in-sensitive to local edits and can irreversibly affect non-edited regions, leaving obvious editing traces. To tackle these problems, we propose a Zero-shot instructiON-guided local image Editing approach, termed ZONE. We first convert the editing intent from the user-provided instruction (e.g., “make his tie blue”) into specific image editing regions through InstructPix2Pix. We then propose a Region-loll scheme for precise image layer extraction from an off-the-shelf segment model. We further develop an edge smoother based on FFT for seamless blending between the layer and the image. Our method allows for arbitrary manipulation of a specific region with a single instruction while preserving the rest. Extensive experiments demonstrate that our Z ONE achieves remarkable local editing results and user-friendliness, outperforming state-of-the-art methods. Code is available at https://github.com/ls1001006/ZONE.

1-bit detectors show impressive performance comparable to their real-valued counterparts when detecting commonly sized objects while exhibiting significant performance degradation on tiny objects. The challenge stems from the fact that high-level features extracted by 1-bit convolutions seem less compelling to reveal the discriminative foreground features. To address these issues, we introduce a Discriminative Feature Refinement method for 1-bit Detectors (DFR-Det), aiming to enhance the discriminative ability of foreground representation for tiny objects in aerial images. This is accomplished by refining the feature representation using an information bottleneck (IB) to achieve a distinctive representation of tiny objects. Specifically, we introduce a new decoder with a foreground mask, aiming to enhance the discriminative ability of high-level features for the target but suppress the background impact. Additionally, our decoder is simple but effective and can be easily mounted on existing detectors without extra burden added to the inference procedure. Extensive experiments on various tiny object detection (TOD) tasks demonstrate DFR-Det’s superiority over state-of-the-art 1-bit detectors. For example, 1-bit FCOS achieved by DFR-Det achieves the 12.8% AP on AI-TOD dataset, approaching the performance of the real-valued counterpart.

The CLIP network excels in various tasks, but struggles with text-visual images i.e., images that contain both text and visual objects; it risks confusing textual and visual representations. To address this issue, we propose MirrorCLIP, a zero-shot framework, which disentangles the image features of CLIP by exploiting the difference in the mirror effect between visual objects and text in the images. Specifically, MirrorCLIP takes both original and flipped images as inputs, comparing their features dimension-wise in the latent space to generate disentangling masks. With disentangling masks, we further design filters to separate textual and visual factors more precisely, and then get disentangled representations. Qualitative experiments using stable diffusion models and class activation mapping (CAM) validate the effectiveness of our disentanglement. Moreover, our proposed MirrorCLIP reduces confusion when encountering text-visual images and achieves a substantial improvement on typographic defense, further demonstrating its superior ability of disentanglement. Our code is available at https://github.com/tcwangbuaa/MirrorCLIP

Semi-supervised learning (SSL), thanks to the significant reduction of data annotation costs, has been an active research topic for large-scale 3D scene understanding. However, the existing SSL-based methods suffer from severe training bias, mainly due to class imbalance and long-tail distributions of the point cloud data. As a result, they lead to a biased prediction for the tail class segmentation. In this paper, we introduce a new decoupling optimization framework, which disentangles feature representation learning and classifier in an alternative optimization manner to shift the bias decision boundary effectively. In particular, we first employ two-round pseudo-label generation to select unlabeled points across head-to-tail classes. We further introduce multi-class imbalanced focus loss to adaptively pay more attention to feature learning across head-to-tail classes. We fix the backbone parameters after feature learning and retrain the classifier using ground-truth points to update its parameters. Extensive experiments demonstrate the effectiveness of our method outperforming previous state-of-the-art methods on both indoor and outdoor 3D point cloud datasets ( i.e. , S3DIS, ScanNet-V2, Semantic3D, and SemanticKITTI) using 1% and 1pt evaluation.

Buildings not only constitute a significant proportion of man-made structures but also serve as a crucial component of geographic information databases, closely linked to human activities. The intelligent interpretation of buildings plays a significant role in urban planning and management, macroeconomic analysis, population dynamics, etc. Remote sensing image building interpretation primarily encompasses building extraction and change detection (CD). However, current methodologies often treat these two tasks as separate entities, thereby failing to leverage shared knowledge. Moreover, the complexity and diversity of remote sensing image scenes pose additional challenges, as most algorithms are designed to model individual small datasets, thus lacking cross-scene generalization. In this article, we propose a comprehensive remote sensing image building understanding model, termed RSBuilding, developed from the perspective of the foundation model. RSBuilding is designed to enhance cross-scene generalization and task universality. Specifically, we extract image features based on the prior knowledge of the foundation model and devise a multilevel feature sampler to augment scale information. To unify task representation and integrate image spatiotemporal clues, we introduce a cross-attention decoder with task prompts. Addressing the current shortage of datasets that incorporate annotations for both tasks, we have developed a federated training strategy to facilitate smooth model convergence even when supervision for some tasks is missing, thereby bolstering the complementarity of different tasks. Our model was trained on a dataset comprising up to 245 000 images and validated on multiple building extraction and CD datasets. The experimental results substantiate that RSBuilding can concurrently handle two structurally distinct tasks and exhibits robust zero-shot generalization capabilities. The code will be made available for open-source access athttps://github.com/Meize0729/RSBuilding.

Industrial sensors capture critical information for intelligent manufacturing maintenance. To promote equipment upgrading and manufacturing processes, intelligent decisions, and information learning play an important role. Although deep learning methods historically obtain excellent results, there is always a tradeoff between fine-tuning existing networks or designing models from scratch for sensor data processing. In this article, we propose the multihead attention self-supervised (MAS) representation model, which is a self-supervised learning-based sensor feature extraction network. To the best of our knowledge, this is the first time a self-supervised contrastive learning method using positive samples that represent multidimensional industry sensor data is being used for anomaly detection. We review alternative data augmentation methods proposed for better-representing sensor sequence data. We use this insight to design a new structure that adapts to the temporal characteristics of the application. We apply our method to a real-world water circulation system that uses a variety of industrial sensors. The effectiveness of the proposed MAS methods is demonstrated.

Online Knowledge Distillation (OKD) is designed to alleviate the dilemma that the high-capacity pre-trained teacher model is not available. However, the existing methods mostly focus on improving the ensemble prediction accuracy from multiple students (a.k.a. branches), which often overlook the homogenization problem that makes student model saturate quickly and hurts the performance. We assume that the intrinsic bottleneck of the homogenization problem comes from the identical branch architecture and coarse ensemble strategy. We propose a novel Adaptive Hierarchy-Branch Fusion framework for Online Knowledge Distillation, termed AHBF-OKD, which designs hierarchical branches and adaptive hierarchy-branch fusion module to boost the model diversity and aggregate complementary knowledge. Specifically, we first introduce hierarchical branch architectures to construct diverse peers by increasing the depth of branches monotonously on the basis of target branch. To effectively transfer knowledge from the most complex branch to the simplest target branch, we propose an adaptive hierarchy-branch fusion module to create hierarchical teacher assistants recursively, which regards the target branch as the smallest teacher assistant. During the training, the teacher assistant from the previous hierarchy is explicitly distilled by the teacher assistant and the branch from the current hierarchy. Thus, the important scores to different branches are effectively and adaptively allocated to reduce the branch homogenization. Extensive experiments demonstrate the effectiveness of AHBF-OKD on different datasets, including CIFAR-10/100 and ImageNet 2012. For example, on ImageNet 2012, the distilled ResNet-18 achieves Top-1 error of 29.28\%, which significantly outperforms the state-of-the-art methods. The source code is available at https://github.com/linruigong965/AHBF.

Binary neural networks (BNNs) have received ever-increasing popularity for their great capability of reducing storage burden as well as quickening inference time. However, there is a severe performance drop compared with {real-valued} networks, due to its intrinsic frequent weight oscillation during training. In this paper, we introduce a Resilient Binary Neural Network (ReBNN) to mitigate the frequent oscillation for better BNNs' training. We identify that the weight oscillation mainly stems from the non-parametric scaling factor. To address this issue, we propose to parameterize the scaling factor and introduce a weighted reconstruction loss to build an adaptive training objective. For the first time, we show that the weight oscillation is controlled by the balanced parameter attached to the reconstruction loss, which provides a theoretical foundation to parameterize it in back propagation. Based on this, we learn our ReBNN by calculating the balanced parameter based on its maximum magnitude, which can effectively mitigate the weight oscillation with a resilient training process. Extensive experiments are conducted upon various network models, such as ResNet and Faster-RCNN for computer vision, as well as BERT for natural language processing. The results demonstrate the overwhelming performance of our ReBNN over prior arts. For example, our ReBNN achieves 66.9% Top-1 accuracy with ResNet-18 backbone on the ImageNet dataset, surpassing existing state-of-the-arts by a significant margin. Our code is open-sourced at https://github.com/SteveTsui/ReBNN.

Image super-resolution (SR) has attracted increasing attention due to its widespread applications. However, current SR methods generally suffer from over-smoothing and artifacts, and most work only with fixed magnifications. This paper introduces an Implicit Diffusion Model (IDM) for high-fidelity continuous image super-resolution. IDM integrates an implicit neural representation and a denoising diffusion model in a unified end-to-end framework, where the implicit neural representation is adopted in the decoding process to learn continuous-resolution representation. Furthermore, we design a scale-adaptive conditioning mechanism that consists of a low-resolution (LR) conditioning network and a scaling factor. The scaling factor regulates the resolution and accordingly modulates the proportion of the LR information and generated features in the final output, which enables the model to accommodate the continuous-resolution requirement. Extensive experiments validate the effectiveness of our IDM and demonstrate its superior performance over prior arts. The source code will be available at https://github.com/Ree1s/IDM.

Continual learning aims to enable a model to incrementally learn knowledge from sequentially arrived data. Previous works adopt the conventional classification architecture, which consists of a feature extractor and a classifier. The feature extractor is shared across sequentially arrived tasks or classes, but one specific group of weights of the classifier corresponding to one new class should be incrementally expanded. Consequently, the parameters of a continual learner gradually increase. Moreover, as the classifier contains all historical arrived classes, a certain size of the memory is usually required to store rehearsal data to mitigate classifier bias and catastrophic forgetting. In this paper, we propose a non-incremental learner, named AttriCLIP, to incrementally extract knowledge of new classes or tasks. Specifically, AttriCLIP is built upon the pre-trained visual-language model CLIP. Its image encoder and text encoder are fixed to extract features from both images and text. Text consists of a category name and a fixed number of learnable parameters which are selected from our designed attribute word bank and serve as attributes. As we compute the visual and textual similarity for classification, AttriCLIP is a non-incremental learner. The attribute prompts, which encode the common knowledge useful for classification, can effectively mitigate the catastrophic forgetting and avoid constructing a replay memory. We evaluate our AttriCLIP and compare it with CLIP-based and previous state-of-the-art continual learning methods in realistic settings with domain-shift and long-sequence learning. The results show that our method performs favorably against previous state-of-the-arts. The implementation code will be available at https://gitee.com/mindspore/models/tree/master/research/cv/AttriCLIP.

Pre-trained vision-language models have inspired much research on few-shot learning. However, with only a few training images, there exist two crucial problems: (1) the visual feature distributions are easily distracted by class-irrelevant information in images, and (2) the alignment between the visual and language feature distributions is difficult. To deal with the distraction problem, we propose a Selective Attack module, which consists of trainable adapters that generate spatial attention maps of images to guide the attacks on class-irrelevant image areas. By messing up these areas, the critical features are captured and the visual distributions of image features are calibrated. To better align the visual and language feature distributions that describe the same object class, we propose a cross-modal distribution alignment module, in which we introduce a vision-language prototype for each class to align the distributions, and adopt the Earth Mover's Distance (EMD) to optimize the prototypes. For efficient computation, the upper bound of EMD is derived. In addition, we propose an augmentation strategy to increase the diversity of the images and the text prompts, which can reduce overfitting to the few-shot training images. Extensive experiments on 11 datasets demonstrate that our method consistently outperforms prior arts in few-shot learning. The implementation code will be available at https://gitee.com/mindspore/models/tree/master/research/cv/SADA.

The recent detection transformer (DETR) has advanced object detection, but its application on resource-constrained devices requires massive computation and memory resources. Quantization stands out as a solution by representing the network in low-bit parameters and operations. However, there is a significant performance drop when performing low-bit quantized DETR (Q-DETR) with existing quantization methods. We find that the bottle-necks of Q-DETR come from the query information distortion through our empirical analyses. This paper addresses this problem based on a distribution rectification distillation (DRD). We formulate our DRD as a bi-level optimization problem, which can be derived by generalizing the information bottleneck (IB) principle to the learning of Q-DETR. At the inner level, we conduct a distribution alignment for the queries to maximize the self-information entropy. At the upper level, we introduce a new foreground-aware query matching scheme to effectively transfer the teacher information to distillation-desired features to minimize the conditional information entropy. Extensive experimental results show that our method performs much better than prior arts. For example, the 4-bit Q-DETR can theoretically accelerate DETR with ResNet-50 backbone by 6.6× and achieve 39.4% AP, with only 2.6% performance gaps than its real-valued counterpart on the COCO dataset11Code: https://github.com/SteveTsui/Q-DETR.

In this paper, we focus on developing knowledge distillation (KD) for compact 3D detectors. We observe that off-the-shelf KD methods manifest their efficacy only when the teacher model and student counterpart share similar intermediate feature representations. This might explain why they are less effective in building extreme-compact 3D detectors where significant representation disparity arises due primarily to the intrinsic sparsity and irregularity in 3D point clouds. This paper presents a novel representation disparity-aware distillation (RDD) method to address the representation disparity issue and reduce performance gap between compact students and over-parameterized teachers. This is accomplished by building our RDD from an innovative perspective of information bottleneck (IB), which can effectively minimize the disparity of proposal region pairs from student and teacher in features and logits. Extensive experiments are performed to demonstrate the superiority of our RDD over existing KD methods. For example, our RDD increases mAP of CP-Voxel-S to 57.1% on nuScenes dataset, which even surpasses teacher performance while taking up only 42% FLOPs.

Vision-language pre-training (VLP) on large-scale datasets has shown premier performance on various downstream tasks. In contrast to plenty of available benchmarks with English corpus, large-scale pre-training datasets and downstream datasets with Chinese corpus remain largely unexplored. In this work, we build a large-scale high-quality Chinese Cross-Modal Benchmark named CCMB for the research community, which contains the currently largest public pre-training dataset Zero and five human-annotated fine-tuning datasets for downstream tasks. Zero contains 250 million images paired with 750 million text descriptions, plus two of the five fine-tuning datasets are also currently the largest ones for Chinese cross-modal downstream tasks. Along with the CCMB, we also develop a VLP framework named R2D2, applying a pre-Ranking + Ranking strategy to learn powerful vision-language representations and a two-way distillation method (i.e., target-guided Distillation and feature-guided Distillation) to further enhance the learning capability. With the Zero and the R2D2 VLP framework, we achieve state-of-the-art performance on twelve downstream datasets from five broad categories of tasks including image-text retrieval, image-text matching, image caption, text-to-image generation, and zero-shot image classification. The datasets, models, and codes are available at https://github.com/yuxie11/R2D2

Denoising diffusion generative models are capable of generating high-quality data, but suffers from the computation-costly generation process, due to a iterative noise estimation using full-precision networks. As an intuitive solution, quantization can significantly reduce the computational and memory consumption by low-bit parameters and operations. However, low-bit noise estimation networks in diffusion models (DMs) remain unexplored yet and perform much worse than the full-precision counterparts as observed in our experimental studies. In this paper, we first identify that the bottlenecks of low-bit quantized DMs come from a large distribution oscillation on activations and accumulated quantization error caused by the multi-step denoising process. To address these issues, we first develop a Timestep-aware Quantization (TaQ) method and a Noise-estimating Mimicking (NeM) scheme for low-bit quantized DMs (Q-DM) to effectively eliminate such oscillation and accumulated error respectively, leading to well-performed low-bit DMs. In this way, we propose an efficient Q-DM to calculate low-bit DMs by considering both training and inference process in the same framework. We evaluate our methods on popular DDPM and DDIM models. Extensive experimental results show that our method achieves a much better performance than the prior arts. For example, the 4-bit Q-DM theoretically accelerates the 1000-step DDPM by 7.8x and achieves a FID score of 5.17, on the unconditional CIFAR-10 dataset.

Binary neural networks (BNNs) have attracted broad research interest due to their efficient storage and computational ability. Nevertheless, a significant challenge of BNNs lies in handling discrete constraints while ensuring bit entropy maximization, which typically makes their weight optimization very difficult. Existing methods relax the learning using the sign function, which simply encodes positive weights into +1s, and -1s otherwise. Alternatively, we formulate an angle alignment objective to constrain the weight binarization to$\lbrace$0,+1$\rbrace$to solve the challenge. In this paper, we show that our weight binarization provides an analytical solution by encoding high-magnitude weights into +1s, and 0 s otherwise. Therefore, a high-quality discrete solution is established in a computationally efficient manner without the sign function. We prove that the learned weights of binarized networks roughly follow a Laplacian distribution that does not allow entropy maximization, and further demonstrate that it can be effectively solved by simply removing the$\ell _{2}$regularization during network training. Our method, dubbed sign-to-magnitude network binarization (SiMaN), is evaluated on CIFAR-10 and ImageNet, demonstrating its superiority over the sign-based state-of-the-arts. Our source code, experimental settings, training logs and binary models are available athttps://github.com/lmbxmu/SiMaN.

Anomaly detection in surveillance videos aims to identify frames where abnormal events happen. Existing approaches assume that the training and testing videos are from the same scene, exhibiting poor generalization performance when encountering an unseen scene. In this paper, we propose a Variational Anomaly Detection Network (VADNet), which is characterized by its high scene-adaptation - it can identify abnormal events in a new scene only via referring to a few normal samples without fine-tuning. Our model embodies two major innovations. First, a novel Variational Normal Inference (VNI) module is proposed to formulate image reconstruction in a conditional variational auto-encoder (CVAE) framework, which learns a probabilistic decision model instead of a traditional deterministic one. Secondly, a Margin Learning Embedding (MLE) module is leveraged to boost the variational inference and aid in distinguishing normal events. We theoretically demonstrate that minimizing the triplet loss in MLE module facilitates maximizing the evidence lower bound (ELBO) of CVAE, which promotes the convergence of VNI. By incorporating variational inference with margin learning, VADNet becomes much more generative that is able to handle the uncertainty caused by the changed scene and limited reference data. Extensive experiments on several datasets demonstrate that the proposed VADNet can adapt to a new scene effectively without fine-tuning and achieve remarkable performance, which outperforms other methods significantly and establishes new state-of-the-art in the case of few-shot scene-adaptive anomaly detection. We believe our method is closer to real-world application due to its strong generalization ability. All codes are released inhttps://github.com/huangxx156/VADNet.

Depth estimation attracts widespread attention in the computer vision community. However, it is still quite difficult to recover an accurate depth map using only one RGB image. We observe a phenomenon that existing methods tend to fail in different cases, caused by differences in network architecture, loss function and so on. In this work, we investigate into the phenomenon and propose to integrate the strengths of multiple weak depth predictor to build a comprehensive and accurate depth predictor, which is critical for many real-world applications, e.g., 3D reconstruction. Specifically, we construct multiple base (weak) depth predictors by utilizing different Transformer-based and convolutional neural network (CNN)-based architectures. Transformer establishes long-range correlation while CNN preserves local information ignored by Transformer due to the spatial inductive bias. Therefore, the coupling of Transformer and CNN contributes to the generation of complementary depth estimates, which are essential to achieve a comprehensive depth predictor. Then, we design mixers to learn from multiple weak predictions and adaptively fuse them into a strong depth estimate. The resultant model, which we refer to as Transformer-assisted depth ensembles (TEDepth). On the standard NYU-Depth-v2 and KITTI datasets, we thoroughly explore how the neural ensembles affect the depth estimation and demonstrate that our TEDepth achieves better results than previous state-of-the-art approaches. To validate the generalizability across cameras, we directly apply the models trained on NYU-Depth-v2 to the SUN RGB-D dataset without any fine-tuning, and the superior results emphasize its strong generalizability.

Modeling the spatiotemporal relationship (STR) of traffic data is important yet challenging for existing graph networks. These methods usually capture features separately in temporal and spatial dimensions or represent the spatiotemporal data by adopting multiple local spatial-temporal graphs. The first kind of method mentioned above is difficult to capture potential temporal-spatial relationships, while the other is limited for long-term feature extraction due to its local receptive field. To handle these issues, the Synchronous Spatio-Temporal grAph Transformer (S2TAT) network is proposed for efficiently modeling the traffic data. The contributions of our method include the following: 1) the nonlocal STR can be synchronously modeled by our integrated attention mechanism and graph convolution in the proposed S2TAT block; 2) the timewise graph convolution and multihead mechanism designed can handle the heterogeneity of data; and 3) we introduce a novel attention-based strategy in the output module, being able to capture more valuable historical information to overcome the shortcoming of conventional average aggregation. Extensive experiments are conducted on PeMS datasets that demonstrate the efficacy of the S2TAT by achieving a top-one accuracy but less computational cost by comparing with the state of the art.

Automatic modulation classification is a challenging and critical task in the field of communication. Deep convolutional networks (ConvNets) have been recently applied in cognitive radio and achieved remarkable performance. However, existing ConvNet-based methods mainly focus on the first-order architecture design, while fail to explore feature correlations of radio signal, which are particularly significant for useful information extraction in low signal-to-noise ratios (SNRs). In this paper, we propose high-order convolutional attention networks (HoCANs) for radio signal expression and feature correlation learning, based on a novel high-order attention mechanism to rescale the convolutional features along channel and sequence dimensions. High-order convolutional layer and covariance matrix after nonlinear transformation are led for tenser filtering with more discriminative representations of radio signals. Experiments have been conducted to validate the superiority of HoCANs which achieve state-of-the-art accuracy for automatic modulation classification on RADIOML 2018.01A dataset.

Energy-based latent variable models (EBLVMs) are more expressive than conventional energy-based models. However, its potential on visual tasks are limited by its training process based on maximum likelihood estimate that requires sampling from two intractable distributions. In this paper, we propose Bi-level doubly variational learning (BiDVL), which is based on a new bi-level optimization framework and two tractable variational distributions to facilitate learning EBLVMs. Particularly, we lead a decoupled EBLVM consisting of a marginal energy-based distribution and a structural posterior to handle the difficulties when learning deep EBLVMs on images. By choosing a symmetric KL divergence in the lower level of our framework, a compact BiDVL for visual tasks can be obtained. Our model achieves impressive image generation performance over related works. It also demonstrates the significant capacity of testing image reconstruction and out-of-distribution detection.

Learning to synthesize data has emerged as a promising direction in zero-shot quantization (ZSQ), which represents neural networks by low-bit integer without accessing any of the real data. In this paper, we observe an interesting phenomenon of intra-class heterogeneity in real data and show that existing methods fail to retain this property in their synthetic images, which causes a limited performance increase. To address this issue, we propose a novel zero-shot quantization method referred to as IntraQ. First, we propose a local object reinforcement that locates the target objects at different scales and positions of the synthetic images. Second, we introduce a marginal distance constraint to form class-related features distributed in a coarse area. Lastly, we devise a soft inception loss which injects a soft prior label to prevent the synthetic images from being over-fitting to a fixed object. Our IntraQ is demonstrated to well retain the intra-class heterogeneity in the synthetic images and also observed to perform state-of-the-art. For example, compared to the advanced ZSQ, our IntraQ obtains 9.17% increase of the top-1 accuracy on ImageNet when all layers of MobileNetV1 are quantized to 4-bit. Code is at https://github.com/zysxmu/IntraQ

Binary neural networks (BNNs) contribute a lot to the efficiency of image classification models. However, in dense predication tasks such as human pose estimation, predictions in different locations are coupled and rely on the extraction of features across entire images. As a result, more robust and adaptive binarization is required to bridge the performance gap between binarized and full precision models. We propose two approaches to conduct image-aware and pixel-aware dynamic binarization in a model for human pose estimation. Firstly, a simplified dynamic thresholding is leveraged in the backbone to determine unique binarization thresholds for each image. Secondly, in the decoder, we decouple binarization for each pixel according to the activations surrounding the pixel. Dynamic filtering modules are proposed to determine a different binarization strategy for each pixel. Compared with the strong baselines, the proposed framework improves 5.2% and 3.6% mAP on the COCO test-dev benchmark for ResNet-18/34 architectures respectively.

Aspect-based Sentiment Classification (ASC) task is a challenge in Natural Language Processing (NLP) and is especially important for fields that require detailed analysis like finance. It aims to identify the sentiment polarity of specific aspects in sentences. In addition to tweets and posts directly related to finance, news from such as restaurants and e-commerce may also indirectly affect its stock prices. In previous approaches, attention-based neural network models were mostly adopted to implicitly connect aspects with opinion words for better aspect representations. However, due to the complexity of language and the presence of multiple aspects in a single sentence, these existing models often confuse connections. To tackle this problem, we propose a model named GAS-CL which encodes syntactical structure into aspect representations and refines it with a contrastive loss. Experiments on several datasets confirm that our approach can have better aspect representations and achieve a significant improvement.

The large pre-trained vision transformers (ViTs) have demonstrated remarkable performance on various visual tasks, but suffer from expensive computational and memory cost problems when deployed on resource-constrained devices. Among the powerful compression approaches, quantization extremely reduces the computation and memory consumption by low-bit parameters and bit-wise operations. However, low-bit ViTs remain largely unexplored and usually suffer from a significant performance drop compared with the real-valued counterparts. In this work, through extensive empirical analysis, we first identify the bottleneck for severe performance drop comes from the information distortion of the low-bit quantized self-attention map. We then develop an information rectification module (IRM) and a distribution guided distillation (DGD) scheme for fully quantized vision transformers (Q-ViT) to effectively eliminate such distortion, leading to a fully quantized ViTs. We evaluate our methods on popular DeiT and Swin backbones. Extensive experimental results show that our method achieves a much better performance than the prior arts. For example, our Q-ViT can theoretically accelerates the ViT-S by 6.14x and achieves about 80.9% Top-1 accuracy, even surpassing the full-precision counterpart by 1.0% on ImageNet dataset. Our codes and models are attached on https://github.com/YanjingLi0202/Q-ViT

Face animation, one of the hottest topics in computer vision, has achieved a promising performance with the help of generative models. However, it remains a critical challenge to generate identity preserving and photo-realistic images due to the sophisticated motion deformation and complex facial detail modeling. To address these problems, we propose a Face Neural Volume Rendering (FNeVR) network to fully explore the potential of 2D motion warping and 3D volume rendering in a unified framework. In FNeVR, we design a 3D Face Volume Rendering (FVR) module to enhance the facial details for image rendering. Specifically, we first extract 3D information with a well designed architecture, and then introduce an orthogonal adaptive ray-sampling module for efficient rendering. We also design a lightweight pose editor, enabling FNeVR to edit the facial pose in a simple yet effective way. Extensive experiments show that our FNeVR obtains the best overall quality and performance on widely used talking-head benchmarks.

Uncertainty estimation in medical image registration enables surgeons to evaluate the operative risk based on the trustworthiness of the registered image data thus of paramount importance for practical clinical applications. Despite the recent promising results obtained with deep unsupervised learning-based registration methods, reasoning about uncertainty of unsupervised registration models remains largely unexplored. In this work, we propose a predictive module to learn the registration and uncertainty in correspondence simultaneously. Our framework introduces empirical randomness and registration error based uncertainty prediction. We systematically assess the performances on two MRI datasets with different ensemble paradigms. Experimental results highlight that our proposed framework significantly improves the registration accuracy and uncertainty compared with the baseline.

Online image hashing has received increasing research attention recently, which processes large-scale data in a streaming fashion to update the hash functions on-the-fly. To this end, most existing works exploit this problem under a supervised setting, i.e., using class labels to boost the hashing performance, which suffers from the defects in both adaptivity and efficiency: First, large amounts of training batches are required to learn up-to-date hash functions, which leads to poor online adaptivity. Second, the training is time-consuming, which contradicts with the core need of online learning. In this paper, a novel supervised online hashing scheme, termed Fast Class-wise Updating for Online Hashing (FCOH), is proposed to address the above two challenges by introducing a novel and efficient inner product operation. To achieve fast online adaptivity, a class-wise updating method is developed to decompose the binary code learning and alternatively renew the hash functions in a class-wise fashion, which well addresses the burden on large amounts of training batches. Quantitatively, such a decomposition further leads to at least 75 percent storage saving. To further achieve online efficiency, we propose a semi-relaxation optimization, which accelerates the online training by treating different binary constraints independently. Without additional constraints and variables, the time complexity is significantly reduced. Such a scheme is also quantitatively shown to well preserve past information during updating hashing functions. We have quantitatively demonstrated that the collective effort of class-wise updating and semi-relaxation optimization provides a superior performance comparing to various state-of-the-art methods, which is verified through extensive experiments on three widely-used datasets.

Gait recognition can be used in person identification and re-identification by itself or in conjunction with other biometrics. Although gait has both spatial and temporal attributes, and it has been observed that decoupling spatial feature and temporal feature can better exploit the gait feature on the fine-grained level. However, the spatial-temporal correlations of gait video signals are also lost in the decoupling process. Direct 3D convolution approaches can retain such correlations, but they also introduce unnecessary interferences. Instead of common 3D convolution solutions, this paper proposes an integration of decoupling process into a 3D convolution framework for cross-view gait recognition. In particular, a novel block consisting of a Parallel-insight Convolution layer integrated with a Spatial-Temporal Dual-Attention (STDA) unit is proposed as the basic block for global spatial-temporal information extraction. Under the guidance of the STDA unit, this block can well integrate spatial-temporal information extracted by two decoupled models and at the same time retain the spatial-temporal correlations. In addition, a Multi-Scale Salient Feature Extractor is proposed to further exploit the fine-grained features through context awareness extension of part-based features and adaptively aggregating the spatial features. Extensive experiments on three popular gait datasets, namely CASIA-B, OULP and OUMVLP, demonstrate that the proposed method outperforms state-of-the-art methods.

In this paper, we find that the conventional convolution operation becomes the bottleneck for extremely efficient binary neural networks (BNNs). To address this issue, we open up a new direction by introducing a reshaped point-wise convolution (RPC) to replace the conventional one to build BNNs. Specifically, we conduct a point-wise convolution after rearranging the spatial information into depth, with which at least$2.25\times $computation reduction can be achieved. Such an efficient RPC allows us to explore more powerful representational capacity of BNNs under a given computation complexity budget. Moreover, we propose to use a balanced activation (BA) to adjust the distribution of the scaled activations after binarization, which enables significant performance improvement of BNNs. After integrating RPC and BA, the proposed network, dubbed as RB-Net, strikes a good trade-off between accuracy and efficiency, achieving superior performance with lower computational cost against the state-of-the-art BNN methods. Specifically, our RB-Net achieves 66.8% Top-1 accuracy with ResNet-18 backbone on ImageNet, exceeding the state-of-the-art Real-to-Binary Net (65.4%) by 1.4% while achieving more than$3\times $reduction (52M vs. 165M) in computational complexity.

Electroencephalography (EEG) is a commonly used clinical approach for the diagnosis of epilepsy which is a life-threatening neurological disorder. Many algorithms have been proposed for the automatic detection of epileptic seizures using traditional machine learning and deep learning. Although deep learning methods have achieved great success in many fields, their performance in EEG analysis and classification is still limited mainly due to the relatively small sizes of available datasets. In this paper, we propose an automatic method for the detection of epileptic seizures based on deep metric learning which is a novel strategy tackling the few-shot problem by mitigating the demand for massive data. First, two one-dimensional convolutional embedding modules are proposed as a deep feature extractor, for single-channel and multichannel EEG signals respectively. Then, a deep metric learning model is detailed along with a stage-wise training strategy. Experiments are conducted on the publicly-available Bonn University dataset which is a benchmark dataset, and the CHB-MIT dataset which is larger and more realistic. Impressive averaged accuracy of 98.60% and specificity of 100% are achieved on the most difficult classification of interictal (subset D) vs ictal (subset E) of the Bonn dataset. On the CHB-MIT dataset, an averaged accuracy of 86.68% and specificity of 93.71% are reached. With the proposed method, automatic and accurate detection of seizures can be performed in real time, and the heavy burden of neurologists can be effectively reduced.

Pedestrian detection in the wild remains a challenging problem especially for scenes containing serious occlusion. In this paper, we propose a novel feature learning method in the deep learning framework, referred to as Feature Calibration Network (FC-Net), to adaptively detect pedestrians under various occlusions. FC-Net is based on the observation that the visible parts of pedestrians are selective and decisive for detection, and is implemented as a self-paced feature learning framework with a self-activation (SA) module and a feature calibration (FC) module. In a new self-activated manner, FC-Net learns features which highlight the visible parts and suppress the occluded parts of pedestrians. The SA module estimates pedestrian activation maps by reusing classifier weights, without any additional parameter involved, therefore resulting in an extremely parsimony model to reinforce the semantics of features, while the FC module calibrates the convolutional features for adaptive pedestrian representation in both pixel-wise and region-based ways. Experiments on CityPersons and Caltech datasets demonstrate that FC-Net improves detection performance on occluded pedestrians up to 10% while maintaining excellent performance on non-occluded instances.

Skeleton-based action recognition has been extensively studied, but it remains an unsolved problem because of the complex variations of skeleton joints in 3-D spatiotemporal space. To handle this issue, we propose a newly temporal-then-spatial recalibration method named memory attention networks (MANs) and deploy MANs using the temporal attention recalibration module (TARM) and spatiotemporal convolution module (STCM). In the TARM, a novel temporal attention mechanism is built based on residual learning to recalibrate frames of skeleton data temporally. In the STCM, the recalibrated sequence is transformed or encoded as the input of CNNs to further model the spatiotemporal information of skeleton sequence. Based on MANs, a new collaborative memory fusion module (CMFM) is proposed to further improve the efficiency, leading to the collaborative MANs (C-MANs), trained with two streams of base MANs. TARM, STCM, and CMFM form a single network seamlessly and enable the whole network to be trained in an end-to-end fashion. Comparing with the state-of-the-art methods, MANs and C-MANs improve the performance significantly and achieve the best results on six data sets for action recognition. The source code has been made publicly available at https://github.com/memory-attention-networks.

Deep encoder-decoders are the model of choice for pixel-level estimation due to their redundant deep architectures. Yet they still suffer from the vanishing supervision information issue that affects convergence because of their overly deep architectures. In this work, we propose and theoretically derive an enhanced deep supervision (EDS) method which improves on conventional deep supervision (DS) by incorporating variance minimization into the optimization. A new structure variance loss is introduced to build a bridge between deep encoder-decoders and variance minimization, and provides a new way to minimize the variance by forcing different intermediate decoding outputs (paths) to reach an agreement. We also design a focal weighting strategy to effectively combine multiple losses in a scale-balanced way, so that the supervision information is sufficiently enforced throughout the encoder-decoders. To evaluate the proposed method on the pixel-level estimation task, a novel multipath residual encoder is proposed and extensive experiments are conducted on four challenging density estimation and crowd counting benchmarks. The experimental results demonstrate the superiority of our EDS over other paradigms, and improved estimation performance is reported using our deeply supervised encoder-decoder.

Modern convolutional neural network (CNN)-based object detectors focus on feature configuration during training but often ignore feature optimization during inference. In this article, we propose a new feature optimization approach to enhance features and suppress background noise in both the training and inference stages. We introduce a generic inference-aware feature filtering (IFF) module that can be easily combined with existing detectors, resulting in our iffDetector. Unlike conventional open-loop feature calculation approaches without feedback, the proposed IFF module performs the closed-loop feature optimization by leveraging high-level semantics to enhance the convolutional features. By applying the Fourier transform to analyze our detector, we prove that the IFF module acts as a negative feedback that can theoretically guarantee the stability of the feature learning. IFF can be fused with CNN-based object detectors in a plug-and-play manner with little computational cost overhead. Experiments on the PASCAL VOC and MS COCO datasets demonstrate that our iffDetector consistently outperforms state-of-the-art methods with significant margins.

Person re-identification (Re-ID) has been promoted by the significant success of convolutional neural networks (CNNs). However, the application of such CNN-based Re-ID methods depends on the tremendous consumption of computation and memory resources, which affects its development on resource-limited devices such as next generation AI chips. As a result, CNN binarization has attracted increasing attention, which leads to binary neural networks (BNNs). In this article, we propose a new BNN-based framework for efficient person Re-ID (BiRe-ID). In this work, we discover that the significant performance drop of binarized models for Re-ID task is caused by the degraded representation capacity of kernels and features. To address the issues, we propose the kernel and feature refinement based on generative adversarial learning (KR-GAL and FR-GAL) to enhance the representation capacity of BNNs. We first introduce an adversarial attention mechanism to refine the binarized kernels based on their real-valued counterparts. Specifically, we introduce a scale factor to restore the scale of 1-bit convolution. And we employ an effective generative adversarial learning method to train the attention-aware scale factor. Furthermore, we introduce a self-supervised generative adversarial network to refine the low-level features using the corresponding high-level semantic information. Extensive experiments demonstrate that our BiRe-ID can be effectively implemented on various mainstream backbones for the Re-ID task. In terms of the performance, our BiRe-ID surpasses existing binarization methods by significant margins, at the level even comparable with the real-valued counterparts. For example, on Market-1501, BiRe-ID achieves 64.0% mAP on ResNet-18 backbone, with an impressive 12.51× speedup in theory and 11.75× storage saving. In particular, the KR-GAL and FR-GAL methods show strong generalization on multiple tasks such as Re-ID, image classification, object detection, and 3D point cloud processing.

Binary Neural Networks (BNNs) have received significant attention due to the memory and computation efficiency recently. However, the considerable accuracy gap between BNNs and their full-precision counterparts hinders BNNs to be deployed to resource-constrained platforms. One of the main reasons for the performance gap can be attributed to the frequent weight flip, which is caused by the misleading weight update in BNNs. To address this issue, we propose a state-aware binary neural network (SA-BNN) equipped with the well designed state-aware gradient. Our SA-BNN is inspired by the observation that the frequent weight flip is more likely to occur, when the gradient magnitude for all quantization states {-1,1} is identical. Accordingly, we propose to employ independent gradient coefficients for different states when updating the weights. Furthermore, we also analyze the effectiveness of the state-aware gradient on suppressing the frequent weight flip problem. Experiments on ImageNet show that the proposed SA-BNN outperforms the current state-of-the-arts (e.g., Bi-Real Net) by more than 3% when using a ResNet architecture. Specifically, we achieve 61.7%, 65.5% and 68.7% Top-1 accuracy with ResNet-18, ResNet-34 and ResNet-50 on ImageNet, respectively.

Ternary neural networks (TNNs) are potential for network acceleration by reducing the full-precision weights in network to ternary ones, e.g., {-1,0,1}. However, existing TNNs are mostly calculated based on rule-of-thumb quantization methods by simply thresholding operations, which causes a significant accuracy loss. In this paper, we introduce a stem-residual framework which provides new insight into Ternary quantization, termed Residual Quantization (TRQ), to achieve more powerful TNNs. Rather than directly thresholding operations, TRQ recursively performs quantization on full-precision weights for a refined reconstruction by combining the binarized stem and residual parts. With such a unique quantization process, TRQ endows the quantizer with high flexibility and precision. Our TRQ is generic, which can be easily extended to multiple bits through recursively encoded residual for a better recognition accuracy. Extensive experimental results demonstrate that the proposed method yields great recognition accuracy while being accelerated.

Panoptic segmentation aims to partition an image to object instances and semantic content for thing and stuff categories, respectively. To date, learning weakly supervised panoptic segmentation (WSPS) with only image-level labels remains unexplored. In this paper, we propose an efficient jointly thing-and-stuff mining (JTSM) framework for WSPS. To this end, we design a novel mask of interest pooling (MoIPool) to extract fixed-size pixel-accurate feature maps of arbitrary-shape segmentations. MoIPool enables a panoptic mining branch to leverage multiple instance learning (MIL) to recognize things and stuff segmentation in a unified manner. We further refine segmentation masks with parallel instance and semantic segmentation branches via self-training, which collaborates the mined masks from panoptic mining with bottom-up object evidence as pseudo-ground-truth labels to improve spatial coherence and contour localization. Experimental results demonstrate the effectiveness of JTSM on PASCAL VOC and MS COCO. As a by-product, we achieve competitive results for weakly supervised object detection and instance segmentation. This work is a first step towards tackling challenge panoptic segmentation task with only image-level labels.

1-bit detectors show great promise for resource-constrained embedded devices but often suffer from a significant performance gap compared with their real-valued counterparts. The primary reason lies in the error during binarization. This paper presents a layer-wise searching (LWS) strategy to generate 1-bit detectors that maintain a performance very close to the original real-valued model. The approach introduces angular and amplitude loss functions to increase detector capacity. At 1-bit layers, it exploits a differentiable binarization search (DBS) to minimize the angular error in a student-teacher framework. We also learn the scale factor by minimizing the amplitude loss in the same student-teacher framework. Extensive experiments show that LWS-Det outperforms state-of-the-art 1-bit detectors by a considerable margin on the PASCAL VOC and COCO datasets. For example, the LWS-Det achieves 1-bit Faster-RCNN with ResNet-34 backbone within 2.0% mAP of its real-valued counterpart on the PASCAL VOC dataset.

The decision-based black-box attack means to craft adversarial examples with only the top-1 label of the victim model available. A common practice is to start from a large perturbation and then iteratively reduce it with a deterministic direction and a random one while keeping it adversarial. The limited information obtained from each query and inefficient direction sampling impede attack efficiency, making it hard to obtain a small enough perturbation within a limited number of queries. To tackle this problem, we propose a novel attack method termed Adaptive History-driven Attack (AHA) which gathers information from all historical queries as the prior for current sampling. Moreover, to balance between the deterministic direction and the random one, we dynamically adjust the coefficient according to the ratio of the actual magnitude reduction to the expected one. Such a strategy improves the success rate of queries during optimization, letting adversarial examples move swiftly along the decision boundary. Our method can also integrate with subspace optimization like dimension reduction to further improve efficiency. Extensive experiments on both ImageNet and CelebA datasets demonstrate that our method achieves at least 24.3% lower magnitude of perturbation on average with the same number of queries. Finally, we prove the practical potential of our method by evaluating it on popular defense methods and a real-world system provided by MEGVII Face++.

Weakly supervised instance segmentation (WSIS) with only image-level labels has recently drawn much attention. To date, bottom-up WSIS methods refine discriminative cues from classifiers with sophisticated multi-stage training procedures, which also suffer from inconsistent object boundaries. And top-down WSIS methods are formulated as cascade detection-to-segmentation pipeline, in which the quality of segmentation learning heavily depends on pseudo masks generated from detectors. In this paper, we propose a unified parallel detection-and-segmentation learning (PDSL) framework to learn instance segmentation with only image-level labels, which draws inspiration from both top-down and bottom-up instance segmentation approaches. The detection module is the same as the typical design of any weakly supervised object detection, while the segmentation module leverages self-supervised learning to model class-agnostic foreground extraction, following by self-training to refine class-specific segmentation. We further design instance-activation correlation module to improve the coherence between detection and segmentation branches. Extensive experiments verify that the proposed method outperforms baselines and achieves the state-of-the-art results on PASCAL VOC and MS COCO.

Differentiable Architecture Search (DARTS) improves the efficiency of architecture search by learning the architecture and network parameters end-to-end. However, the intrinsic relationship between the architecture’s parameters is neglected, leading to a sub-optimal optimization process. The reason lies in the fact that the gradient descent method used in DARTS ignores the coupling relationship of the parameters and therefore degrades the optimization. In this paper, we address this issue by formulating DARTS as a bi-linear optimization problem and introducing an Interactive Differentiable Architecture Search (IDARTS). We first develop a backtracking backpropagation process, which can decouple the relationships of different kinds of parameters and train them in the same framework. The backtracking method coordinates the training of different parameters that fully explore their interaction and optimize training. We present experiments on the CIFAR10 and ImageNet datasets that demonstrate the efficacy of the IDARTS approach by achieving a top-1 accuracy of 76.52% on ImageNet without additional search cost vs. 75.8% with the state-of-the-art PC-DARTS.

Self-supervised learning algorithms that compute depth map from monocular videos have achieved remarkable performance on urban scenes and have been applied extensively. These techniques still face significant challenges, however, when applied directly to endoscopic videos because of the brightness variations from frame to frame and inadequate representation learning during the training phase. Inspired by the optical flow for motion alignment between adjacent frames, we design a AFNet with structural stability loss and residual-based smoothness loss to learn the appearance flow across adjacent frames, which handles the brightness inconsistency issue efficaciously. In addition, we propose a novel self-attention mechanism named feature scaling module to alleviate the inadequate representation learning problem. In a comparison study to the current state-of-the-art self-supervised methods explored for urban videos on the SCARED dataset, the developed model surpasses existing methods by a large margin.

Binary Neural Networks (BNN) are promising machine learning solutions for deployment on resource-limited devices. Recent approaches to training BNNs have produced impressive results, but minimizing the drop in accuracy from full precision networks is still challenging. One reason is that conventional BNNs ignore the uncertainty caused by weights that are near zero, resulting in the instability or frequent flip while learning. In this work, we investigate the intrinsic uncertainty of vanishing near-zero weights, making the training vulnerable to instability. We introduce an uncertainty-aware BNN (UaBNN) by leveraging a new mapping function called certainty-sign (c-sign) to reduce these weights' uncertainties. Our c-sign function is the first to train BNNs with a decreasing uncertainty for binarization. The approach leads to a controlled learning process for BNNs. We also introduce a simple but effective method to measure the uncertainty-based on a Gaussian function. Extensive experiments demonstrate that our method improves multiple BNN methods by maintaining stability of training, and achieves a higher performance over prior arts.

Transformers with remarkable global representation capacities achieve competitive results for visual tasks, but fail to consider high-level local pattern information in input images. In this paper, we present a generic Dual-stream Network (DS-Net) to fully explore the representation capacity of local and global pattern features for image classification. Our DS-Net can simultaneously calculate fine-grained and integrated features and efficiently fuse them. Specifically, we propose an Intra-scale Propagation module to process two different resolutions in each block and an Inter-Scale Alignment module to perform information interaction across features at dual scales. Besides, we also design a Dual-stream FPN (DS-FPN) to further enhance contextual information for downstream dense predictions. Without bells and whistles, the proposed DS-Net outperforms DeiT-Small by 2.4\% in terms of top-1 accuracy on ImageNet-1k and achieves state-of-the-art performance over other Vision Transformers and ResNets. For object detection and instance segmentation, DS-Net-Small respectively outperforms ResNet-50 by 6.4\% and 5.5 \% in terms of mAP on MSCOCO 2017, and surpasses the previous state-of-the-art scheme, which significantly demonstrates its potential to be a general backbone in vision tasks. The code will be released soon.

In recent years, deep learning has dominated progress in the field of medical image analysis. We find however, that the ability of current deep learning approaches to represent the complex geometric structures of many medical images is insufficient. One limitation is that deep learning models require a tremendous amount of data, and it is very difficult to obtain a sufficient amount with the necessary detail. A second limitation is that there are underlying features of these medical images that are well established, but the black-box nature of existing convolutional neural networks (CNNs) do not allow us to exploit them. In this paper, we revisit Gabor filters and introduce a deformable Gabor convolution (DGConv) to expand deep networks interpretability and enable complex spatial variations. The features are learned at deformable sampling locations with adaptive Gabor convolutions to improve representitiveness and robustness to complex objects. The DGConv replaces standard convolutional layers and is easily trained end-to-end, resulting in deformable Gabor feature network (DGFN) with few additional parameters and minimal additional training cost. We introduce DGFN for addressing deep multi-instance multi-label classification on the INbreast dataset for mammograms and on the ChestX-ray14 dataset for pulmonary x-ray images.

In medical image analysis, one limitation of the application of machine learning is the insufficient amount of data with detailed annotation, due primarily to high cost. Another impediment is the domain gap observed between images from different organs and different collections. The differences are even more challenging for the nuclei instance segmentation, where images have significant nuclei stain distribution variations and complex pleomorphisms (sizes and shapes). In this work, we generate style consistent histopathology images for nuclei instance segmentation. We set up a novel instance segmentation framework that integrates a generator and discriminator into the segmentation pipeline with adversarial training to generalize nuclei instances and texture patterns. A segmentation net detects and segments both real nuclei and synthetic nuclei and provides feedback so that the generator can synthesize images that can boost the segmentation performance. Experimental results on three public nuclei datasets indicate that our proposed method outperforms previous nuclei segmentation methods.

Learning-based multi-view stereo (MVS) has recently gained great popularity, which can efficiently infer depth map and reconstruct fine-grained scene geometry. Previous methods calculate the variance of the corresponding pixel pairs to determine whether they are matched mostly based on the pixel-wise measure, which fails to consider the interdependence among pixels and is ineffective on the matching of texture-less or occluded regions. These false matching problems challenge MVS and result in its most failure cases. To address the issues, we introduce a Long-range Attention Network (LANet) to selectively aggregate reference features to each position to capture the long-range interdependence across the entire space. As a result, similar features relate to each other regardless of their distance, propagating more guiding information for the effective match. Furthermore, we introduce a new loss to supervise the intermediate probability volume by constraining its distribution reasonably centered at the true depth. Extensive experiments on large-scale DTU dataset demonstrate that the proposed LANet achieves the new state-of-the-art performance, outperforming previous methods by a large margin. Our method is generic and also achieves comparable results on outdoor Tanks and Temples dataset without any fine-tuning, which validates our method's generalization ability.

Neural architecture search (NAS) has achieved unprecedented performance in various computer vision tasks. However, most existing NAS methods are defected in search efficiency and model generalizability. In this paper, we propose a novel NAS framework, termed MIGO-NAS, with the aim to guarantee the efficiency and generalizability in arbitrary search spaces. On the one hand, we formulate the search space as a multivariate probabilistic distribution, which is then optimized by a novel multivariate information-geometric optimization (MIGO). By approximating the distribution with a sampling, training, and testing pipeline, MIGO guarantees the memory efficiency, training efficiency, and search flexibility. Besides, MIGO is the first time to decrease the estimation error of natural gradient in multivariate distribution. On the other hand, for a set of specific constraints, the neural architectures are generated by a novel dynamic programming network generation (DPNG), which significantly reduces the training cost under various hardware environments. Experiments validate the advantages of our approach over existing methods by establishing a superior accuracy and efficiency i.e., 2.39 test error on CIFAR-10 benchmark and 21.7 on ImageNet benchmark, with only 1.5 GPU hours and 96 GPU hours for searching, respectively. Besides, the searched architectures can be well generalize to computer vision tasks including object detection and semantic segmentation, i.e., 25× FLOPs compression, with 6.4 mAP gain over Pascal VOC dataset, and 29.9× FLOPs compression, with only 1.41 percent performance drop over Cityscapes dataset. The code is publicly available.

Pedestrian trajectory prediction is vital for transportation systems. Generally we can divide pedestrian behavior modeling into two categories, i.e., knowledge-driven and data-driven. The former might bring expert bias, and it sometimes generates unrealistic pedestrian movement due to unnecessary repulsive forces. The latter approach is popular nowadays but most existing neural networks, including fully connected long short-term memory (LSTM) networks, use a 1D vector to model their input and state. The shortcoming is that these works cannot learn spatial information about pedestrians, especially in a dense crowd. To tackle this, we propose to use tensors to represent essential environment features of pedestrians. Accordingly, a convolutional LSTM is designed and deepened to predict spatiotemporal trajectory sequences. As the tensor and convolution can learn better spatiotemporal interactions among pedestrians and environments, experimental results show that the proposed network can estimate more realistic trajectories for a dense crowd in evacuation and counterflow.

Fine-grained visual categorization (FGVC) aims to automatically recognize objects from different sub-ordinate categories. Despite attracting considerable attention from both academia and industry, it remains a challenging task due to subtle visual differences among different classes. Cross-layer feature aggregation and cross-image pairwise learning become prevailing in improving the performance of FGVC by extracting discriminative class-specific features. However, they are still inefficient to fully use the cross-layer information based on the simple aggregation strategy, while existing pairwise learning methods also fail to explore long-range interactions between different images. To address these problems, we propose a novel Alignment Enhancement Network (AENet), including two-level alignments, Cross-layer Alignment (CLA) and Cross-image Alignment (CIA). The CLA module exploits the cross-layer relationship between low-level spatial information and high-level semantic information, which contributes to cross-layer feature aggregation to improve the capacity of feature representation for input images. The new CIA module is further introduced to produce the aligned feature map, which can enhance the relevant information as well as suppress the irrelevant information across the whole spatial region. Our method is based on an underlying assumption that the aligned feature map should be closer to the inputs of CIA when they belong to the same category. Accordingly, we establish Semantic Affinity Loss to supervise the feature alignment within each CIA block. Experimental results on four challenging datasets show that the proposed AENet achieves the state-of-the-art results over prior arts.

Neural architecture search (NAS) can have a significant impact in computer vision by automatically designing optimal neural network architectures for various tasks. A variant, binarized neural architecture search (BNAS), with a search space of binarized convolutions, can produce extremely compressed models. Unfortunately, this area remains largely unexplored. BNAS is more challenging than NAS due to the learning inefficiency caused by optimization requirements and the huge architecture space. To address these issues, we introduce channel sampling and operation space reduction into a differentiable NAS to significantly reduce the cost of searching. This is accomplished through a performance-based strategy used to abandon less potential operations. Two optimization methods for binarized neural networks are used to validate the effectiveness of our BNAS. Extensive experiments demonstrate that the proposed BNAS achieves a performance comparable to NAS on both CIFAR and ImageNet databases. An accuracy of 96.53% vs. 97.22% is achieved on the CIFAR-10 dataset, but with a significantly compressed model, and a 40% faster search than the state-of-the-art PC-DARTS.

Neural network pruning offers a promising prospect to facilitate deploying deep neural networks on resource-limited devices. However, existing methods are still challenged by the training inefficiency and labor cost in pruning designs, due to missing theoretical guidance of non-salient network components. In this paper, we propose a novel filter pruning method by exploring the High Rank of feature maps (HRank). Our HRank is inspired by the discovery that the average rank of multiple feature maps generated by a single filter is always the same, regardless of the number of image batches CNNs receive. Based on HRank, we develop a method that is mathematically formulated to prune filters with low-rank feature maps. The principle behind our pruning is that low-rank feature maps contain less information, and thus pruned results can be easily reproduced. Besides, we experimentally show that weights with high-rank feature maps contain more important information, such that even when a portion is not updated, very little damage would be done to the model performance. Without introducing any additional constraints, HRank leads to significant improvements over the state-of-the-arts in terms of FLOPs and parameters reduction, with similar accuracies. For example, with ResNet-110, we achieve a 58.2%-FLOPs reduction by removing 59.2% of the parameters, with only a small loss of 0.14% in top-1 accuracy on CIFAR-10. With Res-50, we achieve a 43.8%-FLOPs reduction by removing 36.7% of the parameters, with only a loss of 1.17% in the top-1 accuracy on ImageNet. The codes can be available at https://github.com/lmbxmu/HRank.

Conventional learning methods simplify the bilinear model by regarding two intrinsically coupled factors independently, which degrades the optimization procedure. One reason lies in the insufficient training due to the asynchronous gradient descent, which results in vanishing gradients for the coupled variables. In this paper, we introduce a Cogradient Descent algorithm (CoGD) to address the bilinear problem, based on a theoretical framework to coordinate the gradient of hidden variables via a projection function. We solve one variable by considering its coupling relationship with the other, leading to a synchronous gradient descent to facilitate the optimization procedure. Our algorithm is applied to solve problems with one variable under the sparsity constraint, which is widely used in the learning paradigm. We validate our CoGD considering an extensive set of applications including image reconstruction, inpainting, and network pruning. Experiments show that it improves the state-of-the-art by a significant margin.

Dynamic scene classification on portable platforms is extremely challenging due to the contradiction between model complexity and computing resources. To resolve this long-standing dilemma, we propose the compressed attentive multi-granularity network (CAM-Net) in a two-step manner. First, we present a novel AM-Net based on multi-granularity attention units to boost the performance of the full-precision model. It captures and enhances both coarse and fine target-related information. Then, we introduce an efficient binary approximation to AM-Net to improve computing efficiency, leading to CAM-Net. Particularly, a grouping guidance approach is adopted to guide the reconstruction of full-precision weights from binary ones. With this guidance, CAM-Net can significantly reduce memory usage as well as CPU consumption, yet only cause a slight decline in accuracy. Extensive experiments have been conducted on three benchmark datasets, i.e., Maryland, YUPENN ++ and ActivityNet, demonstrating the effectiveness and superiority of the proposed method on scene classification.

Channel pruning is among the predominant approaches to compress deep neural networks. To this end, most existing pruning methods focus on selecting channels (filters) by importance/optimization or regularization based on rule-of-thumb designs, which defects in sub-optimal pruning. In this paper, we propose a new channel pruning method based on artificial bee colony algorithm (ABC), dubbed as ABCPruner, which aims to efficiently find optimal pruned structure, i.e., channel number in each layer, rather than selecting "important" channels as previous works did. To solve the intractably huge combinations of pruned structure for deep networks, we first propose to shrink the combinations where the preserved channels are limited to a specific space, thus the combinations of pruned structure can be significantly reduced. And then, we formulate the search of optimal pruned structure as an optimization problem and integrate the ABC algorithm to solve it in an automatic manner to lessen human interference. ABCPruner has been demonstrated to be more effective, which also enables the fine-tuning to be conducted efficiently in an end-to-end manner. The source codes can be available at https: //github.com/lmbxmu/ABCPruner.

Neural architecture search (NAS) proves to be among the best approaches for many tasks by generating an application-adaptive neural architectures, which are still challenged by high computational cost and memory consumption. At the same time, 1-bit convolutional neural networks (CNNs) with binarized weights and activations show their potential for resource-limited embedded devices. One natural approach is to use 1-bit CNNs to reduce the computation and memory cost of NAS by taking advantage of the strengths of each in a unified framework. To this end, a Child-Parent model is introduced to a differentiable NAS to search the binarized architecture(Child) under the supervision of a full-precision model (Parent). In the search stage, the Child-Parent model uses an indicator generated by the parent and child model accuracy to evaluate the performance and abandon operations with less potential. In the training stage, a kernel level CP loss is introduced to optimize the binarized network. Extensive experiments demonstrate that the proposed CP-NAS achieves a comparable accuracy with traditional NAS on both the CIFAR and ImageNet databases. It achieves an accuracy of 95.27% on CIFAR-10, 64.3% on ImageNet with binarized weights and activations, and a 30% faster search than prior arts.

Binary Neural Network (BNN) shows its predominance in reducing the complexity of deep neural networks. However, it suffers severe performance degradation. One of the major impediments is the large quantization error between the full-precision weight vector and its binary vector. Previous works focus on compensating for the norm gap while leaving the angular bias hardly touched. In this paper, for the first time, we explore the influence of angular bias on the quantization error and then introduce a Rotated Binary Neural Network (RBNN), which considers the angle alignment between the full-precision weight vector and its binarized version. At the beginning of each training epoch, we propose to rotate the full-precision weight vector to its binary vector to reduce the angular bias. To avoid the high complexity of learning a large rotation matrix, we further introduce a bi-rotation formulation that learns two smaller rotation matrices. In the training stage, we devise an adjustable rotated weight vector for binarization to escape the potential local optimum. Our rotation leads to around 50% weight flips which maximize the information gain. Finally, we propose a training-aware approximation of the sign function for the gradient backward. Experiments on CIFAR-10 and ImageNet demonstrate the superiorities of RBNN over many state-of-the-arts. Our source code, experimental settings, training logs and binary models are available at https://github.com/lmbxmu/RBNN.

Generative models have been successfully used for anomaly detection, which however need a large number of parameters and computation overheads, especially when training spatial and temporal networks in the same framework. In this paper, we introduce a novel network architecture, Orthogonal Features Fusion Network (OFF-Net), to solve the anomaly detection problem. We show that the convolutional feature maps used for generating future frames are orthogonal with each other, which can improve representation capacity of generative models and strengthen temporal connections between adjacent images. We lead a simple but effective module easily mounted on convolutional neural networks (CNNs) with negligible additional parameters added, which can replace the widely-used optical flow network a nd significantly improve th e performance for anomaly detection. Extensive experiment results demonstrate the effectiveness of OFF-Net that we outperform the state-of-the-art model 1.7% in terms of AUC. We save around 85M-space parameters compared with the prevailing prior arts using optical flow network without comprising the performance.

Current studies have shown the effectiveness of long short-term memory network (LSTM) for skeleton-based human action recognition in capturing temporal and spatial features of the skeleton sequence. Nevertheless, it still remains challenging for LSTM to extract the latent structural dependency among nodes. In this paper, we introduce a new long-short graph memory network (LSGM) to improve the capability of LSTM to model the skeleton sequence - a type of graph data. Our proposed LSGM can learn high-level temporal-spatial features end-to-end, enabling LSTM to extract the spatial information that is neglected but intrinsic to the skeleton graph data. To improve the discriminative ability of the temporal and spatial module, we use a calibration module termed as graph temporal-spatial calibration (GTSC) to calibrate the learned temporal-spatial features. By integrating the two modules into the same framework, we obtain a stronger generalization capability in processing dynamic graph data and achieve a significant performance improvement on the NTU and SYSU dataset. Experimental results have validated the effectiveness of our proposed LSGM+GTSC model in extracting temporal and spatial information from dynamic graph data.1

Video segmentation is a task of partitioning pixels that exhibit homogeneous appearance and motion into coherent spatial-temporal groups, which is still challenging for aerial applications. In this paper, a principled combination of superpixel labeling priors and Markov random field (S-MRF) is proposed for aerial video segmentation. The proposed approach has several contributions: 1) we develop a metadata-based global projection model with coordinate transformation to estimate motion information between frames; 2) the superpixel labeling priors from previous frames are incorporated into the segmentation of the current frame, leading to a highly efficient probabilistic label propagation algorithm; and 3) we perform an MRF optimization on the initial segments with propagated labeling priors to improve the temporal coherency. In addition, a new video dataset is collected and will be made publicly available to evaluate the performance of aerial video segmentation algorithms. The experimental results show that the proposed approach outperforms the state-of-the-art video segmentation methods.

Building assessment is highly prioritized during rescue operations and damage relief after hurricane disasters. Although machine learning has made remarkable improvement in building damage classification, it remains challenging because classifiers must be trained using a massive amount of labeled data. Furthermore, data labeling is labor intensive, costly, and unavailable after a disaster. To address this issue, we propose an unsupervised domain adaptation method with aligned discriminative and representative features (ADRF), which leverage a substantial amount of labeled data of relevant disaster scenes for new classification tasks. The remote sensing imageries of different disasters are collected using different sensors, viewpoints, times, even at various places. Compared with the public datasets used in the domain adaptation community, the remote sensing imageries are more complicated which exhibit characteristics of lower discrimination between categories and higher diversity within categories. As a result, pursuing domain invariance is a huge challenge. To achieve this goal, we build a framework with ADRF to improve the discriminative and representative capability of the extracted features to facilitate the classification task. The ADRF framework consists of three pipelines: a classifier for the labeled data of the source domain and one autoencoder each for the source and target domains. The latent variables of autoencoders are forced to observe unit Gaussian distributions by minimizing the maximum mean discrepancy (MMD), whereas the marginal distributions of both domains are aligned via the MMD. As a case study, two challenging transfer tasks using the hurricane Sandy, Maria, and Irma datasets are investigated. Experimental results demonstrate that ADRF achieves overall accuracy of 71.6% and 84.1% in the transfer tasks from dataset Sandy to dataset Maria and dataset Irma, respectively.

Small object tracking becomes an increasingly important task, which however has been largely unexplored in computer vision. The great challenges stem from the facts that: 1) small objects show extreme vague and variable appearances, and 2) they tend to be lost easier as compared to normal-sized ones due to the shaking of lens. In this paper, we propose a novel aggregation signature suitable for small object tracking, especially aiming for the challenge of sudden and large drift. We make three-fold contributions in this work. First, technically, we propose a new descriptor, named aggregation signature, based on saliency, able to represent highly distinctive features for small objects. Second, theoretically, we prove that the proposed signature matches the foreground object more accurately with a high probability. Third, experimentally, the aggregation signature achieves a high performance on multiple datasets, outperforming the state-of-the-art methods by large margins. Moreover, we contribute with two newly collected benchmark datasets, i.e., small90 and small112, for visually small object tracking. The datasets will be available in https://github.com/bczhangbczhang/.

Representation learning is a fundamental but challenging problem, especially when the distribution of data is unknown. In this paper, we propose a new representation learning method, named Structure Transfer Machine (STM), which enables feature learning process to converge at the representation expectation in a probabilistic way. We theoretically show that such an expected value of the representation (mean) is achievable if the manifold structure can be transferred from the data space to the feature space. The resulting structure regularization term, named manifold loss, is incorporated into the loss function of the typical deep learning pipeline. The STM architecture is constructed to enforce the learned deep representation to satisfy the intrinsic manifold structure from the data, which results in robust features that suit various application scenarios, such as digit recognition, image classification and object tracking. Compared with state-of-the-art CNN architectures, we achieve better results on several commonly used public benchmarks.

Recent advances of person re-identification have well advocated the usage of human body cues to boost performance. However, most existing methods still retain on exploiting a relatively coarse-grained local information. Such information may include redundant backgrounds that are sensitive to the apparently similar persons when facing challenging scenarios like complex poses, inaccurate detection, occlusion and misalignment. In this paper we propose a novel Fine-Grained Spatial Alignment Model (FGSAM) to mine fine-grained local information to handle the aforementioned challenge effectively. In particular, we first design a pose resolve net with channel parse blocks (CPB) to extract pose information in pixel-level. This network allows the proposed model to be robust to complex pose variations while suppressing the redundant backgrounds caused by inaccurate detection and occlusion. Given the extracted pose information, a locally reinforced alignment mode is further proposed to address the misalignment problem between different local parts by considering different local parts along with attribute information in a fine-grained way. Finally, a focal triplet loss is designed to effectively train the entire model, which imposes a constraint on the intra-class and an adaptively weight adjustment mechanism to handle the hard sample problem. Extensive evaluations and analysis on Market1501, DukeMTMC-reid and PETA datasets demonstrate the effectiveness of FGSAM in coping with the problems of misalignment, occlusion and complex poses.

Cell classification, especially that of white blood cells, plays a very important role in the field of diagnosis and control of major diseases. Compared to traditional optical microscopic imaging, hyperspectral imagery, combined with both spatial and spectral information, provides more wealthy information for recognizing cells. In this paper, a novel blood cell classification framework, which combines a modulated Gabor wavelet and deep convolutional neural network (CNN) kernels, named as MGCNN, is proposed based on medical hyperspectral imaging. For each convolutional layer, multi-scale and orientation Gabor operators are taken dot product with initial CNN kernels. The essence is to transform the convolutional kernels into the frequency domain to learn features. By combining characteristics of Gabor wavelets, the features learned by modulated kernels at different frequencies and orientations are more representative and discriminative. Experimental results demonstrate that the proposed model can achieve better classification performance than traditional CNNs and widely used support vector machine approaches, especially as training small-sample-size situations.

Pedestrian detection in the wild remains a challenging problem especially when the scene contains significant occlusion and/or low resolution of the pedestrians to be detected. Existing methods are unable to adapt to these difficult cases while maintaining acceptable performance. In this paper we propose a novel feature learning model, referred to as CircleNet, to achieve feature adaptation by mimicking the process humans looking at low resolution and occluded objects: focusing on it again, at a finer scale, if the object can not be identified clearly for the first time. CircleNet is implemented as a set of feature pyramids and uses weight sharing path augmentation for better feature fusion. It targets at reciprocating feature adaptation and iterative object detection using multiple top-down and bottom-up pathways. To take full advantage of the feature adaptation capability in CircleNet, we design an instance decomposition training strategy to focus on detecting pedestrian instances of various resolutions and different occlusion levels in each cycle. Specifically, CircleNet implements feature ensemble with the idea of hard negative boosting in an end-to-end manner. Experiments on two pedestrian detection datasets, Caltech and CityPersons, show that CircleNet improves the performance of occluded and low-resolution pedestrians with significant margins while maintaining good performance on normal instances.

Crowd counting is a popular topic with widespread applications. Currently, the biggest challenge to crowd counting is large-scale variation in objects. In this article, we focus on overcoming this challenge by proposing a novel Attentive Encoder-Decoder Network (AEDN), which is supervised on multiple feature scales to conduct crowd counting via density estimation. This work has three main contributions. First, we augment the traditional encoder-decoder architecture with our proposed residual attention blocks, which, beyond skip-connected encoded features, further extend the decoded features with attentive features. AEDN is better at establishing long-range dependencies between the encoder and decoder, therefore promoting more effective fusion of multi-scale features for handling scale-variations. Second, we design a new KL-divergence-based distribution loss to supervise the scale-aware structural differences between two density maps, which complements the pixel-isolated MSE loss and better optimizes AEDN to generate high-quality density maps. Third, we adopt a multi-scale supervision scheme, such that multiple KL divergences and MSE losses are deployed at all decoding stages, providing more thorough supervisions for different feature scales. Extensive experimental results on four public datasets, including ShanghaiTech Part A, ShanghaiTech Part B, UCF-CC-50, and UCF-QNRF, reveal the superiority and efficacy of the proposed method, which outperforms most state-of-the-art competitors.

The advancement of deep convolutional neural networks (DCNNs) has driven significant improvement in the accuracy of recognition systems for many computer vision tasks. However, their practical applications are often restricted in resource-constrained environments. In this paper, we introduce projection convolutional neural networks (PCNNs) with a discrete back propagation via projection (DBPP) to improve the performance of binarized neural networks (BNNs). The contributions of our paper include: 1) for the first time, the projection function is exploited to efficiently solve the discrete back propagation problem, which leads to a new highly compressed CNNs (termed PCNNs); 2) by exploiting multiple projections, we learn a set of diverse quantized kernels that compress the full-precision kernels in a more efficient way than those proposed previously; 3) PCNNs achieve the best classification performance compared to other state-ofthe-art BNNs on the ImageNet and CIFAR datasets.

Recent advances on fine-grained image retrieval prefer learning convolutional neural network (CNN) with specific fullyconnect layer designed loss function for discriminative feature representation. Essentially, such loss should establish a robust metric to efficiently distinguish high-dimensional features within and outside fine-grained categories. To this end, the existing loss functions are defected in two aspects: (a) The feature relationship is encoded inside the training batch. Such a local scope leads to low accuracy. (b) The error is established by the mean square, which needs pairwise distance computation in training set and results in low efficiency. In this paper, we propose a novel metric learning scheme, termed Normalize-Scale Layer and Decorrelated Global Centralized Ranking Loss, which achieves extremely efficient and discriminative learning, i.e., 5× speedup over triplet loss and 12% recall boost on CARS196. Our method originates from the classic softmax loss, which has a global structure but does not directly optimize the distance metric as well as the inter/intra class distance. We tackle this issue through a hypersphere layer and a global centralized ranking loss with a pairwise decorrelated learning. In particular, we first propose a Normalize-Scale Layer to eliminate the gap between metric distance (for measuring distance in retrieval) and dot product (for dimension reduction in classification). Second, the relationship between features is encoded under a global centralized ranking loss, which targets at optimizing metric distance globally and accelerating learning procedure. Finally, the centers are further decorrelated by Gram-Schmidt process, leading to extreme efficiency (with 20 epochs in training procedure) and discriminability in feature learning. We have conducted quantitative evaluations on two fine-grained retrieval benchmark. The superior performance demonstrates the merits of the proposed approach over the state-of-the-arts.

In stochastic gradient descent (SGD) and its variants, the optimized gradient estimators may be as expensive to compute as the true gradient in many scenarios. This paper introduces a calibrated stochastic gradient descent (CSGD) algorithm for deep neural network optimization. A theorem is developed to prove that an unbiased estimator for the network variables can be obtained in a probabilistic way based on the Lipschitz hypothesis. Our work is significantly distinct from existing gradient optimization methods, by providing a theoretical framework for unbiased variable estimation in the deep learning paradigm to optimize the model parameter calculation. In particular, we develop a generic gradient calibration layer which can be easily used to build convolutional neural networks (CNNs). Experimental results demonstrate that CNNs with our CSGD optimization scheme can improve the stateof-the-art performance for natural image classification, digit recognition, ImageNet object classification, and object detection tasks. This work opens new research directions for developing more efficient SGD updates and analyzing the backpropagation algorithm.

Crowd counting has recently attracted increasing interest in computer vision but remains a challenging problem. In this paper, we propose a trellis encoder-decoder network (TEDnet) for crowd counting, which focuses on generating high-quality density estimation maps. The major contributions are four-fold. First, we develop a new trellis architecture that incorporates multiple decoding paths to hierarchically aggregate features at different encoding stages, which improves the representative capability of convolutional features for large variations in objects. Second, we employ dense skip connections interleaved across paths to facilitate sufficient multi-scale feature fusions, which also helps TEDnet to absorb the supervision information. Third, we propose a new combinatorial loss to enforce similarities in local coherence and spatial correlation between maps. By distributedly imposing this combinatorial loss on intermediate outputs, TEDnet can improve the back-propagation process and alleviate the gradient vanishing problem. Finally, on four widely-used benchmarks, our TEDnet achieves the best overall performance in terms of both density map quality and counting accuracy, with an improvement up to 14% in MAE metric. These results validate the effectiveness of TEDnet for crowd counting.

Compressing convolutional neural networks (CNNs) has received ever-increasing research focus. However, most existing CNN compression methods do not interpret their inherent structures to distinguish the implicit redundancy. In this paper, we investigate the problem of CNN compression from a novel interpretable perspective. The relationship between the input feature maps and 2D kernels is revealed in a theoretical framework, based on which a kernel sparsity and entropy (KSE) indicator is proposed to quantitate the feature map importance in a feature-agnostic manner to guide model compression. Kernel clustering is further conducted based on the KSE indicator to accomplish high-precision CNN compression. KSE is capable of simultaneously compressing each layer in an efficient way, which is significantly faster compared to previous data-driven feature map pruning methods. We comprehensively evaluate the compression and speedup of the proposed method on CIFAR-10, SVHN and ImageNet 2012. Our method demonstrates superior performance gains over previous ones. In particular, it achieves 4.7× FLOPs reduction and 2.9× compression on ResNet-50 with only a top-5 accuracy drop of 0.35% on ImageNet 2012, which significantly outperforms state-of-the-art methods.

Structured pruning of filters or neurons has received increased focus for compressing convolutional neural networks. Most existing methods rely on multi-stage optimizations in a layer-wise manner for iteratively pruning and retraining which may not be optimal and may be computation intensive. Besides, these methods are designed for pruning a specific structure, such as filter or block structures without jointly pruning heterogeneous structures. In this paper, we propose an effective structured pruning approach that jointly prunes filters as well as other structures in an end-to-end manner. To accomplish this, we first introduce a soft mask to scale the output of these structures by defining a new objective function with sparsity regularization to align the output of baseline and network with this mask. We then effectively solve the optimization problem by generative adversarial learning (GAL), which learns a sparse soft mask in a label-free and an end-to-end manner. By forcing more scale factors in the soft mask to zero, the fast iterative shrinkage-thresholding algorithm (FISTA) can be leveraged to fast and reliably remove the corresponding structures. Extensive experiments demonstrate the effectiveness of GAL on different datasets, including MNIST, CIFAR-10 and ImageNet ILSVRC 2012. For example, on ImageNet ILSVRC 2012, the pruned ResNet-50 achieves 10.88% Top-5 error and results in a factor of 3.7x speedup. This significantly outperforms state-of-the-art methods.

The rapidly decreasing computation and memory cost has recently driven the success of many applications in the field of deep learning. Practical applications of deep learning in resource-limited hardware, such as embedded devices and smart phones, however, remain challenging. For binary convolutional networks, the reason lies in the degraded representation caused by binarizing full-precision filters. To address this problem, we propose new circulant filters (CiFs) and a circulant binary convolution (CBConv) to enhance the capacity of binarized convolutional features via our circulant back propagation (CBP). The CiFs can be easily incorporated into existing deep convolutional neural networks (DCNNs), which leads to new Circulant Binary Convolutional Networks (CBCNs). Extensive experiments confirm that the performance gap between the 1-bit and full-precision DCNNs is minimized by increasing the filter diversity, which further increases the representational ability in our networks. Our experiments on ImageNet show that CBCNs achieve 61.4% top-1 accuracy with ResNet18. Compared to the state-of-the-art such as XNOR, CBCNs can achieve up to 10% higher top-1 accuracy with more powerful representational ability.

Deep learning models have shown their vulnerabilities to universal adversarial perturbations (UAP), which are quasi-imperceptible. Compared to the conventional supervised UAPs that suffer from the knowledge of training data, the data-independent unsupervised UAPs are more applicable. Existing unsupervised methods fail to take advantage of the model uncertainty to produce robust perturbations. In this paper, we propose a new unsupervised universal adversarial perturbation method, termed as Prior Driven Uncertainty Approximation (PD-UA), to generate a robust UAP by fully exploiting the model uncertainty at each network layer. Specifically, a Monte Carlo sampling method is deployed to activate more neurons to increase the model uncertainty for a better adversarial perturbation. Thereafter, a textural bias prior to revealing a statistical uncertainty is proposed, which helps to improve the attacking performance. The UAP is crafted by the stochastic gradient descent algorithm with a boosted momentum optimizer, and a Laplacian pyramid frequency model is finally used to maintain the statistical uncertainty. Extensive experiments demonstrate that our method achieves well attacking performances on the ImageNet validation set, and significantly improves the fooling rate compared with the state-of-the-art methods.

Deep convolutional neural networks (DCNNs) have dominated the recent developments in computer vision through making various record-breaking models. However, it is still a great challenge to achieve powerful DCNNs in resource-limited environments, such as on embedded devices and smart phones. Researchers have realized that 1-bit CNNs can be one feasible solution to resolve the issue; however, they are baffled by the inferior performance compared to the full-precision DCNNs. In this paper, we propose a novel approach, called Bayesian optimized 1-bit CNNs (denoted as BONNs), taking the advantage of Bayesian learning, a well-established strategy for hard problems, to significantly improve the performance of extreme 1-bit CNNs. We incorporate the prior distributions of full-precision kernels and features into the Bayesian framework to construct 1-bit CNNs in an end-to-end manner, which have not been considered in any previous related methods. The Bayesian losses are achieved with a theoretical support to optimize the network simultaneously in both continuous and discrete spaces, aggregating different losses jointly to improve the model capacity. Extensive experiments on the ImageNet and CIFAR datasets show that BONNs achieve the best classification performance compared to state-of-the-art 1-bit CNNs.

Architectures obtained by Neural Architecture Search (NAS) have achieved highly competitive performance in various computer vision tasks. However, the prohibitive computation demand of forward-backward propagation in deep neural networks and searching algorithms makes it difficult to apply NAS in practice. In this paper, we propose a Multinomial Distribution Learning for extremely effective NAS, which considers the search space as a joint multinomial distribution, i.e., the operation between two nodes is sampled from this distribution, and the optimal network structure is obtained by the operations with the most likely probability in this distribution. Therefore, NAS can be transformed to a multinomial distribution learning problem, i.e., the distribution is optimized to have a high expectation of the performance. Besides, a hypothesis that the performance ranking is consistent in every training epoch is proposed and demonstrated to further accelerate the learning process. Experiments on CIFAR-10 and ImageNet demonstrate the effectiveness of our method. On CIFAR-10, the structure searched by our method achieves 2.55% test error, while being 6.0× (only 4 GPU hours on GTX1080Ti) faster compared with state-of-the-art NAS algorithms. On ImageNet, our model achieves 74% top1 accuracy under MobileNet settings (MobileNet V1/V2), while being 1.2× faster with measured GPU latency. Test code with pre-trained models are available at https: //github.com/tanglang96/MDENAS.

This paper introduces an Image Based Visual Servoing (IBVS) system to track a ground target from a flying UAV. We implement a tracking system via two deep neural networks; approaching network and chasing network. In the approaching network, the UAV-target relative distance is estimated through the tracked target state to drive the UAV to a desired distance. In the chasing network, the tracked target orientation is used in resolving the ambiguity between the UAV's yawing and lateral movements and aligning the UAV with the tracked target heading to ease the tracking task. The detection accuracy for both networks has been validated and a series of real flight data, which has been conducted on ground targets and persons showed the proposed system efficiency in target tracking and following.

Vehicle detection and recognition in remote sensing images are challenging, especially when only limited training data are available to accommodate various target categories. In this paper, we introduce a novel coarse-to-fine framework, which decomposes vehicle detection into segmentation-based vehicle localization and generalized zero-shot vehicle classification. Particularly, the proposed framework can well handle the problem of generalized zero-shot vehicle detection, which is challenging due to the requirement of recognizing vehicles that are even unseen during training. Specifically, a hierarchical DeepLab v3 model is proposed in the framework, which fully exploits fine-grained features to locate the target on a pixel-wise level, then recognizes vehicles in a coarse-grained manner. Additionally, the hierarchical DeepLab v3 model is beneficially compatible to combine the generalized zero-shot recognition. To the best of our knowledge, there is no publically available dataset to test comparative methods, we therefore construct a new dataset to fill this gap of evaluation. The experimental results show that the proposed framework yields promising results on the imperative yet difficult task of zero-shot vehicle detection and recognition.

Deep convolutional neural networks (DCNN) with manifold embedding have achieved considerable attention in computer vision. However, prior arts are usually based on the neighborhood-based graph modeling only the pairwise relationship between two samples, which fail to fully capture intra-class variations and thus suffer from severe performance loss for noisy data. While such intra-class variations can be well captured via sophisticated hypergraph structure, we are motivated and lead a hypergraph induced Convolutional Manifold Network (H-CMN) to significantly improve the representation capacity of DCNN for the complex data. Specifically, two innovative designs are provides: 1) our manifold preserving method is implemented based on a mini-batch, which can be efficiently plugged into the existing DCNN training pipelines and be scalable for large datasets; 2) a robust hypergraph is built for each mini-batch, which not only offers a strong robustness against typical noise, but also captures the variances from multiple features. Extensive experiments on the image classification task on large benchmarking datasets demonstrate that our model achieves much better performance than the state-of-the-art

Binarized convolutional neural networks (BCNNs) are widely used to improve memory and computation efficiency of deep convolutional neural networks (DCNNs) for mobile and AI chips based applications. However, current BCNNs are not able to fully explore their corresponding full-precision models, causing a significant performance gap between them. In this paper, we propose rectified binary convolutional networks (RBCNs), towards optimized BCNNs, by combining full-precision kernels and feature maps to rectify the binarization process in a unified framework. In particular, we use a GAN to train the 1-bit binary network with the guidance of its corresponding full-precision model, which significantly improves the performance of BCNNs. The rectified convolutional layers are generic and flexible, and can be easily incorporated into existing DCNNs such as WideResNets and ResNets. Extensive experiments demonstrate the superior performance of the proposed RBCNs over state-of-the-art BCNNs. In particular, our method shows strong generalization on the object tracking task.

Face anti-spoofing detection is critical to guarantee the security of biometric face recognition systems. Despite extensive advances in facial anti-spoofing based on single-model image, little work has been devoted to multi-modal anti-spoofing, which is however widely encountered in real-world scenarios. Following the recent progress, this paper mainly focuses on multi-modal face anti-spoofing and aims to solve the following two challenges: (1) how to effectively fuse multi-modal information; and (2) how to effectively learn distinguishable features despite single cross-entropy loss. We propose a novel Multi-modal Multi-layer Fusion Convolutional Neural Network (mmfCNN), which targets at finding a discriminative model for recognizing the subtle differences between live and spoof faces. The mmfCNN can fully use different information provided by diverse modalities, which is based on a weight-adaptation aggregation approach. Specifically, we utilize a multi-layer fusion model to further aggregate the features from different layers, which fuses the low-, mid- and high-level information from different modalities in a unified framework. Moreover, a novel Average Binary Center (ABC) loss is proposed to maximize the dissimilarity between the features of live and spoof faces, which helps to stabilize the training to generate a robust and discriminative model. Extensive experiments conducted on the CISIA-SURF and 3DMAD datasets verify the significance and generalization capability of the proposed method for the face anti-spoofing task. Code is available at: https://github.com/SkyKuang/Face-anti-spoofing.

Despite the exciting progress in image captioning, generating diverse captions for a given image remains as an open problem. Existing methods typically apply generative models such as Variational Auto-Encoder to diversify the captions, which however neglect two key factors of diverse expression, i.e., the lexical diversity and the syntactic diversity. To model these two inherent diversities in image captioning, we propose a Variational Structured Semantic Inferring model (termed VSSI-cap) executed in a novel structured encoder-inferer-decoder schema. VSSI-cap mainly innovates in a novel structure, i.e., Variational Multi-modal Inferring tree (termed VarMI-tree). In particular, conditioned on the visual-textual features from the encoder, the VarMI-tree models the lexical and syntactic diversities by inferring their latent variables (with variations) in an approximate posterior inference guided by a visual semantic prior. Then, a reconstruction loss and the posterior-prior KL-divergence are jointly estimated to optimize the VSSI-cap model. Finally, diverse captions are generated upon the visual features and the latent variables from this structured encoder-inferer-decoder model. Experiments on the benchmark dataset show that the proposed VSSI-cap achieves significant improvements over the state-of-the-arts.

Correlation filter (CF) trackers have achieved outstanding performance in visual object tracking tasks, in which the cosine mask plays an essential role in alleviating boundary effects caused by the circular assumption. However, the cosine mask imposes a larger weight on its center position, which greatly affects CF trackers, that is, their performance will drop significantly if a bad starting point happens to occur. To address the above issue, we propose a target adaptive image signature (TaiS) model to refine the starting point in each frame for CF trackers. Specifically, we incorporate the target priori into the image signature to build a target-specific saliency map, and iteratively refine the starting point with a closed-form solution during the tracking process. As a result, our TaiS is able to find a better starting point close to the center of targets; more importantly, it is independent of specific CF trackers and can efficiently improve their performance. Experiments on two benchmark datasets, i.e., OTB100 and UAV123, demonstrate that our TaiS consistently achieves high performance and updates the state of the arts in visual tracking. The source code of our approach will be made publicly available.

This paper provides a new perspective to understand CNNs based on the Taylor expansion, leading to new Taylor Convolutional Networks (TaylorNets) for image classification. We introduce a principled combination of the high frequency information (i.e., detailed information) and low frequency information in the end-to-end TaylorNets, based on a nonlinear combination of the convolutional feature maps. The steerable module developed in TaylorNets is generic, which can be easily integrated into well-known deep architectures and learned within the same pipeline of the back propagation algorithm, yielding a higher representation capacity for CNNs. Extensive experimental results demonstrate the super capability of our TaylorNets which improve widely used CNNs architectures, such as conventional CNNs and ResNet, in terms of object classification accuracy on well-known benchmarks. The code will be publicly available.

Object tracking is challenged by the varying appearances of targets and the real‐time requirement. Siamese regression trackers, being one of the most popular tracking paradigms, excel in efficiency but suffer at adaptability to cope with appearance variations. To improve their adaptability, the authors propose a new adaptive Siamese (ASiam) tracker, which integrates a novel adversarial template generation module and a motion‐based failure recovery module. The template generation module exploits the temporal coherence and evolution of target appearance variations encoded in preceding tracklets and then generates an adaptive target template online which approximates the varying target in the coming frame. This generation module is optimised via adversarial learning to achieve accurate appearance prediction and sharp template quality. The generated template, together with a search region, are fed into a Siamese tracking backbone to compute an appearance response map via dense similarity computation in a sliding‐window way. At frames where the Siamese tracking fails, the failure recovery module is invoked to perform deep frame differencing motion detection to provide a motion response map. By fusing different response maps, the drifted tracker can be re‐calibrated. Extensive experiments on the OTB2013, OTB2015, and VOT2016 datasets prove the accuracy and efficiency of the proposed tracker.

Ensemble learning is one of the hottest topics in machine learning. In this letter, we develop a convolutional attention in ensemble (CAE) method, which, for the first time, introduces attention-based weighting scheme into ensemble learning. The knowledge contained in base classifiers is transferred into the final classifier, by which the base classifier with a higher performance could be given much more attention. In particular, we employ convolutional attention models to develop an efficient ensemble classifier for image classification. Our CAE can leverage the representation capacity of convolutional neural networks to enhance the performance of ensemble classifiers. We apply our method to remote sensing image classification tasks, which achieves much better performance than the state of the arts.

Dynamic scene classification has been extensively studied in computer vision due to its widespread applications. The key to dynamic scene classification lies in jointly characterizing spatial appearance and temporal dynamics to achieve informative representation, which remains an outstanding task in the literature. In this paper, we propose a unified framework to extract spatial and temporal features for dynamic scene representation. More specifically, we deploy two variants of deep convolutional neural networks to encode spatial appearance and short-term dynamics into short-term deep features (STDF). Based on STDF, we propose using the autoregressive moving average model to extract long-term frequency features (LTFF). By combining STDF and LTFF, we establish the long-short-term feature (LSTF) representations of dynamic scenes. The LSTF characterizes both spatial and temporal patterns of dynamic scenes for comprehensive and information representation that enables more accurate classification. Extensive experiments on three-dynamic scene classification benchmarks have shown that the proposed LSTF achieves high performance and substantially surpasses the state-of-the-art methods.

Flying bird detection has recently attracted increasing attention in computer vision, which becomes an urgent task with the opening up of the low-altitude airspace. However, compared to conventional object detection tasks, it is much more challenging to trap flying birds in aerial videos due to small target sizes, complex backgrounds of great variations and disturbances of bird-like objects. In this paper, we propose a unified framework termed glance-and-stare detection (GSD) to trap flying birds in aerial videos. The GSD is inspired by the fact that human beings first glance at the whole image and then stare at the areas where the suspected object is most likely to appear until the confirmation is obtained. Specifically, we propose the zooming-in algorithm to generate region proposals for accurate localization of flying birds; to represent region proposal sequences of different lengths, we propose adaptive deep spatio-temporal features by leveraging the strength of 3D convolutional neural networks, based on which classification is conducted to achieve final detection. In contrast to conventional methods, the GSD enables localization and classification to be conducted jointly in an alternating iterative way, which mutually enhances each other to improve their performance. In order to validate the proposed GSD algorithm, we build flying bird data sets including images and videos, which provide new benchmarks for evaluation of flying bird detection systems. Experiments on the data sets demonstrate that the GSD can achieve high detection accuracy and largely outperform the state-of-the-art detection methods.

As one of the fundamental research topics in remote sensing image analysis, hyperspectral image (HSI) classification has been extensively studied so far. However, how to discriminatively learn a low-dimensional feature space, in which the mapped features have small within-class scatter and big between-class separation, is still a challenging problem. To address this issue, this paper proposes an effective framework, named compact and discriminative stacked autoencoder (CDSAE), for HSI classification. The proposed CDSAE framework comprises two stages with different optimization objectives, which can learn discriminative low-dimensional feature mappings and train an effective classifier progressively. First, we impose a local Fisher discriminant regularization on each hidden layer of stacked autoencoder (SAE) to train discriminative SAE (DSAE) by minimizing reconstruction error. This stage can learn feature mappings, in which the pixels from the same land-cover class are mapped as nearly as possible and the pixels from different land-cover categories are separated by a large margin. Second, we learn an effective classifier and meanwhile update DSAE with a local Fisher discriminant regularization being embedded on the top of feature representations. Moreover, to learn a compact DSAE with as small number of hidden neurons as possible, we impose a diversity regularization on the hidden neurons of DSAE to balance the feature dimensionality and the feature representation capability. The experimental results on three widely-used HSI data sets and comprehensive comparisons with existing methods demonstrate that our proposed method is effective.

While intrinsic data structure in subspace provides useful information for visual recognition, it has not yet been well studied in deep feature learning for action recognition. In this paper, we introduce a new spatio-temporal manifold network (STMN) that leverages data manifold structures to regularize deep action feature learning, aiming at simultaneously minimizing the intra-class variations of learned deep features and alleviating the over-fitting problem. To this end, the manifold prior is imposed from the top layer of a convolutional neural network (CNN), and is propagated across convolutional layers during forward-backward propagation. The observed correspondence of manifold structures in the data space and feature space validates that the manifold priori can be transferred across CNN layers. STMN theoretically recasts the problem of transferring the data structure prior into the deep learning architectures as a projection over the manifold via an embedding method, which can be easily solved by an Alternating Direction Method of Multipliers and Backward Propagation (ADMM-BP) algorithm. STMN is generic in the sense that it can be plugged into various backbone architectures to learn more discriminative representation for action recognition. Extensive experimental results show that our method achieves comparable or even better performance as compared with the state-of-the-art approaches on four benchmark datasets.

Video classification has been extensively researched in computer vision due to its wide spread applications. However, it remains an outstanding task because of the great challenges in effective spatial-temporal feature extraction and efficient classification with high-dimensional video representations. To address these challenges, in this paper, we propose an end-to-end learning framework called deep ensemble machine (DEM) for video classification. Specifically, to establish effective spatio-temporal features, we propose using two deep convolutional neural networks (CNNs), i.e., vision and graphics group and C3-D to extract heterogeneous spatial and temporal features for complementary representations. To achieve efficient classification, we propose ensemble learning based on random projections aiming to transform high-dimensional features into a set of lower dimensional compact features in subspaces; an ensemble of classifiers is trained on the subspaces and combined with a weighting layer during the backpropagation. To further enhance the performance, we introduce rectified linear encoding (RLE) inspired from error-correcting output coding to encode the initial outputs of classifiers, followed by a softmax layer to produce the final classification results. DEM combines the strengths of deep CNNs and ensemble learning, which establishes a new end-to-end learning architecture for more accurate and efficient video classification. We show the great effectiveness of DEM by extensive experiments on four data sets for diverse video classification tasks including action recognition and dynamic scene classification. Results have shown that DEM achieves high performance on all tasks with an improvement of up to 13% on CIFAR10 data set over the baseline model.

Despite great effectiveness of very deep and wide Convolutional Neural Networks (CNNs) in various computer vision tasks, the significant cost in terms of storage requirement of such networks impedes the deployment on computationally limited devices. In this paper, we propose new modulated convolutional networks (MCNs) to improve the portability of CNNs via binarized filters. In MCNs, we propose a new loss function which considers the filter loss, center loss and softmax loss in an end-to-end framework. We first introduce modulation filters (M-Filters) to recover the unbinarized filters, which leads to a new architecture to calculate the network model. The convolution operation is further approximated by considering intra-class compactness in the loss function. As a result, our MCNs can reduce the size of required storage space of convolutional filters by a factor of 32, in contrast to the full-precision model, while achieving much better performances than state-of-the-art binarized models. Most importantly, MCNs achieve a comparable performance to the full-precision Resnets and WideResnets. The code will be available publicly soon.

Accelerating convolutional neural networks has recently received ever-increasing research focus. Among various approaches proposed in the literature, filter pruning has been regarded as a promising solution, which is due to its advantage in significant speedup and memory reduction of both network model and intermediate feature maps. To this end, most approaches tend to prune filters in a layer-wise fixed manner, which is incapable to dynamically recover the previously removed filter, as well as jointly optimize the pruned network across layers. In this paper, we propose a novel global & dynamic pruning (GDP) scheme to prune redundant filters for CNN acceleration. In particular, GDP first globally prunes the unsalient filters across all layers by proposing a global discriminative function based on prior knowledge of filters. Second, it dynamically updates the filter saliency all over the pruned sparse network, and then recover the mistakenly pruned filter, followed by a retraining phase to improve the model accuracy. Specially, we effectively solve the corresponding non-convex optimization problem of the proposed GDP via stochastic gradient descent with greedy alternative updating. Extensive experiments show that, comparing to the state-of-the-art filter pruning methods, the proposed approach achieves superior performance to accelerate several cutting-edge CNNs on the ILSVRC 2012 benchmark.

Despite its great success, matrix factorization based cross-modality hashing suffers from two problems: 1) there is no engagement between feature learning and binarization; and 2) most existing methods impose the relaxation strategy by discarding the discrete constraints when learning the hash function, which usually yields suboptimal solutions. In this paper, we propose a novel multimodal hashing framework, referred as Unsupervised Deep Cross-Modal Hashing (UDCMH), for multimodal data search in a self-taught manner via integrating deep learning and matrix factorization with binary latent factor models. On one hand, our unsupervised deep learning framework enables the feature learning to be jointly optimized with the binarization. On the other hand, the hashing system based on the binary latent factor models can generate unified binary codes by solving a discrete-constrained objective function directly with no need for a relaxation step. Moreover, novel Laplacian constraints are incorporated into the objective function, which allow to preserve not only the nearest neighbors that are commonly considered in the literature but also the farthest neighbors of data, even if the semantic labels are not available. Extensive experiments on multiple datasets highlight the superiority of the proposed framework over several state-of-the-art baselines.

Skeleton-based action recognition task is entangled with complex spatio-temporal variations of skeleton joints, and remains challenging for Recurrent Neural Networks (RNNs). In this work, we propose a temporal-then-spatial recalibration scheme to alleviate such complex variations, resulting in an end-to-end Memory Attention Networks (MANs) which consist of a Temporal Attention Recalibration Module (TARM) and a Spatio-Temporal Convolution Module (STCM). Specifically, the TARM is deployed in a residual learning module that employs a novel attention learning network to recalibrate the temporal attention of frames in a skeleton sequence. The STCM treats the attention calibrated skeleton joint sequences as images and leverages the Convolution Neural Networks (CNNs) to further model the spatial and temporal information of skeleton data. These two modules (TARM and STCM) seamlessly form a single network architecture that can be trained in an end-to-end fashion. MANs significantly boost the performance of skeleton-based action recognition and achieve the best results on four challenging benchmark datasets: NTU RGB+D, HDM05, SYSU-3D and UT-Kinect.

Steerable properties dominate the design of traditional filters, e.g., Gabor filters, and endow features the capability of dealing with spatial transformations. However, such excellent properties have not been well explored in the popular deep convolutional neural networks (DCNNs). In this paper, we propose a new deep model, termed Gabor Convolutional Networks (GCNs or Gabor CNNs), which incorporates Gabor filters into DCNNs to enhance the resistance of deep learned features to the orientation and scale changes. By only manipulating the basic element of DCNNs based on Gabor filters, i.e., the convolution operator, GCNs can be easily implemented and are compatible with any popular deep learning architecture. Experimental results demonstrate the super capability of our algorithm in recognizing objects, where the scale and rotation changes occur frequently. The proposed GCNs have much fewer learnable network parameters, and thus is easier to train with an endtoend pipeline. The source code will be here1.

In steerable filters, a filter of arbitrary orientation can be generated by a linear combination of a set of "basis filters." Steerable properties dominate the design of the traditional filters, e.g., Gabor filters and endow features the capability of handling spatial transformations. However, such properties have not yet been well explored in the deep convolutional neural networks (DCNNs). In this paper, we develop a new deep model, namely, Gabor convolutional networks (GCNs or Gabor CNNs), with Gabor filters incorporated into DCNNs such that the robustness of learned features against the orientation and scale changes can be reinforced. By manipulating the basic element of DCNNs, i.e., the convolution operator, based on Gabor filters, GCNs can be easily implemented and are readily compatible with any popular deep learning architecture. We carry out extensive experiments to demonstrate the promising performance of our GCNs framework, and the results show its superiority in recognizing objects, especially when the scale and rotation changes take place frequently. Moreover, the proposed GCNs have much fewer network parameters to be learned and can effectively reduce the training complexity of the network, leading to a more compact deep learning model while still maintaining a high feature representation capacity. The source code can be found at https://github.com/bczhangbczhang.

Correlation filters are special classifiers designed for shift-invariant object recognition, which are robust to pattern distortions. The recent literature shows that combining a set of sub-filters trained based on a single or a small group of images obtains the best performance. The idea is equivalent to estimating variable distribution based on the data sampling (bagging), which can be interpreted as finding solutions (variable distribution approximation) directly from sampled data space. However, this methodology fails to account for the variations existed in the data. In this paper, we introduce an intermediate step-solution sampling-after the data sampling step to form a subspace, in which an optimal solution can be estimated. More specifically, we propose a new method, named latent constrained correlation filters (LCCF), by mapping the correlation filters to a given latent subspace, and develop a new learning framework in the latent subspace that embeds distribution-related constraints into the original problem. To solve the optimization problem, we introduce a subspace-based alternating direction method of multipliers, which is proven to converge at the saddle point. Our approach is successfully applied to three different tasks, including eye localization, car detection, and object tracking. Extensive experiments demonstrate that LCCF outperforms the state-of-the-art methods.11.

Online multi-object tracking (MOT) in an intelligent vehicle platform aims at locating the surrounding objects in real time, which remains far from being solved in complex scenarios, due to various motion patterns of tracked objects and severe occlusions caused by cluttered background or other objects. In this paper, we establish a unified online MOT framework for complex scenarios that employs a hierarchical model to improve the solution of data association, termed hierarchical MOT (HMOT). Incorporating the multiple Gaussians uncertainty theory into the individual motion model for each target followed by imposing interaction constraint to re-associate the tracklets with lower confidence leads our algorithm to achieve accurate multi-object tracking. With such a model, individual objects are not only more precisely associated across frames, but also dynamically constrained with each other in a global manner. Experiments on challenging data sets verify the performance of the proposed HMOT approach over the other state-of-the-art MOT methods.

Binary code learning has been emerging topic in large-scale cross-modality retrieval recently. It aims to map features from multiple modalities into a common Hamming space, where the cross-modality similarity can be approximated efficiently via Hamming distance. To this end, most existing works learn binary codes directly from data instances in multiple modalities, which preserve both intra-and inter-modal similarities respectively. Few methods consider to preserve the fusion similarity among multi-modal instances instead, which can explicitly capture their heterogeneous correlation in cross-modality retrieval. In this paper, we propose a hashing scheme, termed Fusion Similarity Hashing (FSH), which explicitly embeds the graph-based fusion similarity across modalities into a common Hamming space. Inspired by the fusion by diffusion, our core idea is to construct an undirected asymmetric graph to model the fusion similarity among different modalities, upon which a graph hashing scheme with alternating optimization is introduced to learn binary codes that embeds such fusion similarity. Quantitative evaluations on three widely used benchmarks, i.e., UCI Handwritten Digit, MIR-Flickr25K and NUS-WIDE, demonstrate that the proposed FSH approach can achieve superior performance over the state-of-the-art methods.

In this paper, a self-learning approach is proposed towards solving scene-specific pedestrian detection problem without any human annotation involved. The self-learning approach is deployed as progressive steps of object discovery, object enforcement, and label propagation. In the learning procedure, object locations in each frame are treated as latent variables that are solved with a progressive latent model (PLM). Compared with conventional latent models, the proposed PLM incorporates a spatial regularization term to reduce ambiguities in object proposals and to enforce object localization, and also a graph-based label propagation to discover harder instances in adjacent frames. With the difference of convex (DC) objective functions, PLM can be efficiently optimized with a concave-convex programming and thus guaranteeing the stability of self-learning. Extensive experiments demonstrate that even without annotation the proposed self-learning approach outperforms weakly supervised learning approaches, while achieving comparable performance with transfer learning and fully supervised approaches.

In this paper, we present a new strategy for modeling the motion of local patches for single-object tracking that can be seamlessly applied to most part-based trackers in the literature. The proposed adaptive local movement modeling method is able to model the local movement distribution of the image patches defining the object to track and the reliability of each image patch. Given the output of a base tracking algorithm, a Gaussian mixture model (GMM) is first used to model the distribution of the movement of local patches relative to the center of gravity of the tracked object. Then, the GMM is combined with the chosen base tracker in a boosting framework, which gives an efficient integrated scheme for the tracking task. This provides a robust procedure to detect outliers in the local motion of the patches. The algorithm is highly configurable with the possibility to change the number of local patches used for tracking and to adapt to the variations of the tracked object. The extensive tracking results on standard data sets show that equipping state-of-the-art trackers with our technique remarkably improves their performance.

Human action recognition is an important yet challenging task. This paper presents a low-cost descriptor called 3D histograms of texture (3DHoTs) to extract discriminant features from a sequence of depth maps. 3DHoTs are derived from projecting depth frames onto three orthogonal Cartesian planes, i.e., the frontal, side, and top planes, and thus compactly characterize the salient information of a specific action, on which texture features are calculated to represent the action. Besides this fast feature descriptor, a new multi-class boosting classifier (MBC) is also proposed to efficiently exploit different kinds of features in a unified framework for action classification. Compared with the existing boosting frameworks, we add a new multi-class constraint into the objective function, which helps to maintain a better margin distribution by maximizing the mean of margin, whereas still minimizing the variance of margin. Experiments on the MSRAction3D, MSRGesture3D, MSRActivity3D, and UTD-MHAD data sets demonstrate that the proposed system combining 3DHoTs and MBC is superior to the state of the art.

The kernelized correlation filter (KCF) is one of the state-of-the-art object trackers. However, it does not reasonably model the distribution of correlation response during tracking process, which might cause the drifting problem, especially when targets undergo significant appearance changes due to occlusion, camera shaking, and/or deformation. In this paper, we propose an output constraint transfer (OCT) method that by modeling the distribution of correlation response in a Bayesian optimization framework is able to mitigate the drifting problem. OCT builds upon the reasonable assumption that the correlation response to the target image follows a Gaussian distribution, which we exploit to select training samples and reduce model uncertainty. OCT is rooted in a new theory which transfers data distribution to a constraint of the optimized variable, leading to an efficient framework to calculate correlation filters. Extensive experiments on a commonly used tracking benchmark show that the proposed method significantly improves KCF, and achieves better performance than other state-of-the-art trackers. To encourage further developments, the source code is made available.

Human re-identification is still a challenging task due to the human pose and illumination variations. Nowadays, surveillance cameras with high frame rate are capable of capturing several consecutive frames from each person. Multi-shot images provide richer information of the target person compared to a single-shot image. They, however, produce a high cost of information redundancy which may degrade the performance of re-identification systems. In this paper, we propose a novel framework that combines sparse coding and manifold constraints to extract discriminative information from multi-shot images of one pedestrian for person re-identification across a set of non-overlapped surveillance cameras. The evaluation over two standard multi-shot datasets shows very competitive accuracy of our framework against the state-of-the-art.

The fact that image data samples lie on a manifold has been successfully exploited in many learning and inference problems. In this paper we leverage the specific structure of data in order to improve recognition accuracies in general recognition tasks. In particular we propose a novel framework that allows to embed manifold priors into sparse representation-based classification (SRC) approaches. We also show that manifold constraints can be transferred from the data to the optimized variables if these are linearly correlated. Using this new insight, we define an efficient alternating direction method of multipliers (ADMM) that can consistently integrate the manifold constraints during the optimization process. This is based on the property that we can recast the problem as the projection over the manifold via a linear embedding method based on the Geodesic distance. The proposed approach is successfully applied on face, digit, action and objects recognition showing a consistently increase on performance when compared to the state of the art.

In this paper we present a novel strategy for modelling the motion of local patches for single object tracking that can be seamlessly applied to most part-based trackers in the literature. The proposed Adaptive Local Movement Modelling (ALMM) method is able to model the local spatial distribution of the image patches defining the object to track and the reliability of each image patch. Given the output of a base tracking algorithm, a Gaussian Mixture Model (GMM) is first used to model the distribution of the movement of local patches relative to the gravity center of the tracked object. Then, the GMM is combined with the base tracker in a boosting framework, which gives a novel integrated boosting classifier for the tracking task. This provides a robust procedure to detect outliers in the local motion of the patches. The algorithm is highly configurable with the possibility to change the number of local patches used for tracking and to adapt to the variations of the tracked object. Tracking results on standard datasets show that equipping state-of-the-art trackers with our tehcnique remarkably improves their performance.

UNLABELLED: The previously known works studying the generalization ability of support vector machine classification (SVMC) algorithm are usually based on the assumption of independent and identically distributed samples. In this paper, we go far beyond this classical framework by studying the generalization ability of SVMC based on uniformly ergodic Markov chain (u.e.M.c.) samples. We analyze the excess misclassification error of SVMC based on u.e.M.c. samples, and obtain the optimal learning rate of SVMC for u.e.M.c. SAMPLES: We also introduce a new Markov sampling algorithm for SVMC to generate u.e.M.c. samples from given dataset, and present the numerical studies on the learning performance of SVMC based on Markov sampling for benchmark datasets. The numerical studies show that the SVMC based on Markov sampling not only has better generalization ability as the number of training samples are bigger, but also the classifiers based on Markov sampling are sparsity when the size of dataset is bigger with regard to the input dimension.

This paper presents a non-linear pre-processing method based on Similarity Michaelis-Menten law (SMML) for face recognition. Similarity Michaelis-Menten law can be used to explain visual sensitivity in the vertebrate retina. We preprocess input images using SMML, and then employ Local Binary Pattern (LBP) for face feature extraction. Advantages of SMML include improvement of light adaption, noise effect, detection right rate, robustness and efficiency, which inspire us exploit it for face pre-processing descriptor for the first time in the field of face recognition. And the parameters of SMML are spatiotemporally and locally estimated by the input image itself employing Sobel, which shows its advantages for face recognition. Extensive experiments clearly demonstrate the superiority of our method over the ones which only use LBP on FERET database in many aspects including the robustness against different facial expressions, lighting and aging of the subjects.

This paper introduces a new BeiHang (BH) Keystroke Dynamics Database for testing and evaluation of biometric approaches. Different from the existing keystroke dynamics researches which solely rely on laboratory experiments, the developed database is collected from a real commercialized system and thus is more comprehensive and more faithful to human behavior. Moreover, our database comes with ready-to-use benchmark results of three keystroke dynamics methods, Nearest Neighbor classifier, Gaussian Model and One-Class Support Vector Machine. Both the database and benchmark results are open to the public and provide a significant experimental platform for international researchers in the keystroke dynamics area.

This paper proposes a new method named Sparse Regression Analysis (SRA) for object representation and recognition. In SRA, ℓ1-norm minimization is combined with regression analysis to represent the input signal. The discriminative ability of SRA derives from the fact that the subset which most compactly expresses the input signal is activated in the regression analysis. To achieve a further improvement, Kernelized SRA (KSRA) is developed to make a nonlinear extension of SRA. The experiments are conducted on both palmprint and face recognition, which show that the proposed methods achieve a much better performance than sparse representation classifier, principal component analysis, and linear discriminant analysis.

This paper proposes a novel kernel similarity modeling of texture pattern flow (KSM-TPF) for background modeling and motion detection in complex and dynamic environments. The texture pattern flow encodes the binary pattern changes in both spatial and temporal neighborhoods. The integral histogram of texture pattern flow is employed to extract the discriminative features from the input videos. Different from existing uniform threshold based motion detection approaches which are only effective for simple background, the kernel similarity modeling is proposed to produce an adaptive threshold for complex background. The adaptive threshold is computed from the mean and variance of an extended Gaussian mixture model. The proposed KSM-TPF approach incorporates machine learning method with feature extraction method in a homogenous way. Experimental results on the publicly available video sequences demonstrate that the proposed approach provides an effective and efficient way for background modeling and motion detection.

This paper proposes a new learning method, which integrates feature selection with classifier construction for human detection via solving three optimization models. Firstly, the method trains a series of weak-classifiers by the proposed L1-norm Minimization Learning (LML) and min-max penalty function models. Secondly, the proposed method selects the weak-classifiers by using the integer optimization model to construct a strong classifier. The L1-norm minimization and integer optimization models aim to find the minimal VC-dimension for weak and strong classifiers respectively. Finally, the method constructs a cascade of LML (CLML) classifier to reach higher detection rates and efficiency. Histograms of Oriented Gradients features of variable-size blocks (v-HOG) are employed as human representation to verify the proposed method. Experiments conducted on INRIA human test set show more superior detection rates and speed than state-of-the-art methods.

In recent years, sparse representation originating from signal compressed sensing theory has attracted increasing interest in computer vision research community. However, to our best knowledge, no previous work utilizes L1-norm minimization for human detection. In this paper we develop a novel human detection system based on L1-norm Minimization Learning (LML) method. The method is on the observation that a human object can be represented by a few features from a large feature set (sparse representation). And the sparse representation can be learned from the training samples by exploiting the L1-norm Minimization principle, which can also be called feature selection procedure. This procedure enables the feature representation more concise and more adaptive to object occlusion and deformation. After that a classifier is constructed by linearly weighting features and comparing the result with a calculated threshold. Experiments on two datasets validate the effectiveness and efficiency of the proposed method.

In this paper, we propose an approach for fast pedestrian detection in images. Inspired by the histogram of oriented gradient (HOG) features, a set of multi-scale orientation (MSO) features are proposed as the feature representation. The features are extracted on square image blocks of various sizes (called units), containing coarse and fine features in which coarse ones are the unit orientations and fine ones are the pixel orientation histograms of the unit. A cascade of Adaboost is employed to train classifiers on the coarse features, aiming to high detection speed. A greedy searching algorithm is employed to select fine features, which are input into SVMs to train the fine classifiers, aiming to high detection accuracy. Experiments report that our approach obtains state-of-art results with 12.4 times faster than the SVM+HOG method.

Face recognition using micropattern representation has recently received much attention in the computer vision and pattern recognition community. Previous researches demonstrated that micropattern representation based on Gabor features achieves better performance than its direct usage on gray-level images. This paper conducts a comparative performance evaluation of micropattern representations on four forms of Gabor features for face recognition. Three evaluation rules are proposed and observed for a fair comparison. To reduce the high feature dimensionality problem, uniform quantization is used to partition the spatial histograms. The experimental results reveal that: 1) micropattern representation based on Gabor magnitude features outperforms the other three representations, and the performances of the other three are comparable; and 2) micropattern representation based on the combination of Gabor magnitude and phase features performs the best.

This paper proposes a novel high-order local pattern descriptor, local derivative pattern (LDP), for face recognition. LDP is a general framework to encode directional pattern features based on local derivative variations. The n(th)-order LDP is proposed to encode the (n-1)(th) -order local derivative direction variations, which can capture more detailed information than the first-order local pattern used in local binary pattern (LBP). Different from LBP encoding the relationship between the central point and its neighbors, the LDP templates extract high-order local information by encoding various distinctive spatial relationships contained in a given local region. Both gray-level images and Gabor feature images are used to evaluate the comparative performances of LDP and LBP. Extensive experimental results on FERET, CAS-PEAL, CMU-PIE, Extended Yale B, and FRGC databases show that the high-order LDP consistently performs much better than LBP for both face identification and face verification under various conditions.

Dynamic scenes (e.g. waving trees, ripples in water, illumination changes, camera jitters etc.) challenge many traditional background subtraction methods. In this paper, we present a novel background subtraction approach for dynamic scenes, in which the background is modeled in a multi-resolution framework. First, for each level of the pyramid, we run an independent mixture of Gaussians Models (GMM) that outputs a background subtraction map. Second, these background subtraction maps are combined via AND operator to finally get a more robust and accurate background subtraction map. This is a natural fusion because the original resolution and low resolution images have complementary strengths, which original resolution image contains rich information and low resolution image is insensitive to the noises and the small movement of dynamic scene. Experimental result shows that this real-time algorithm is able to detect moving objects accurately even in dynamic scenes.

In this paper, a novel gradient Gabor (GGabor) filter is proposed to extract multi-scale and multi-orientation features to represent and classify faces. Gradient Gabor combines the derivative of Gaussian functions and the harmonic functions to capture the features in both spatial and frequency domains to deliver orientation and scale information. The spatial positions are combined into Gaussian derivatives which allows it to provide more stable information. An efficient Kernel Fisher analysis method is proposed to find multiple subspaces based on both GGabor magnitude and phase features, which is a local kernel mapping method to capture the structure information in faces. Experiments on two face databases, FRGC Version 1 and FRGC Version 2, are conducted to compare the performances of the Gabor and GGabor features, which show that GGabor can also be a powerful tool to model faces, and the Efficient Kernel Fisher classifier can improve the efficiency of the original kernel fisher method.

This paper presents a new Sobel-LBP, an extension of existing Local Binary Pattern (LBP), for facial image representation. The face image is filtered by Sobel operator to enhance the edge information. Sobel-LBP feature distributions are then extracted and concatenated into a spatial histogram to be used as a face descriptor. The proposed method is compared with the original LBP on both gray-level images and Gabor real and imaginary features for face recognition. The experimental results indicate that Sobel-LBP provides a significantly better performance than LBP under various conditions.

This paper proposes a novel texture pattern flow (TPF) for complex background modeling and motion detection. The pattern flow is proposed to encode the binary pattern changes among the neighborhoods in the space-time domain. To model the distribution of the TPF, the TPF integral histograms are used to extract the discriminative features to represent the input video. Experimental results on the public videos testify the effectiveness of the proposed method in comparison to LBP and GMM based background modeling methods.

A novel object descriptor, histogram of Gabor phase pattern (HGPP), is proposed for robust face recognition. In HGPP, the quadrant-bit codes are first extracted from faces based on the Gabor transformation. Global Gabor phase pattern (GGPP) and local Gabor phase pattern (LGPP) are then proposed to encode the phase variations. GGPP captures the variations derived from the orientation changing of Gabor wavelet at a given scale (frequency), while LGPP encodes the local neighborhood variations by using a novel local XOR pattern (LXP) operator. They are both divided into the nonoverlapping rectangular regions, from which spatial histograms are extracted and concatenated into an extended histogram feature to represent the original image. Finally, the recognition is performed by using the nearest-neighbor classifier with histogram intersection as the similarity measurement. The features of HGPP lie in two aspects: 1) HGPP can describe the general face images robustly without the training procedure; 2) HGPP encodes the Gabor phase information, while most previous face recognition methods exploit the Gabor magnitude information. In addition, Fisher separation criterion is further used to improve the performance of HGPP by weighing the subregions of the image according to their discriminative powers. The proposed methods are successfully applied to face recognition, and the experiment results on the large-scale FERET and CAS-PEAL databases show that the proposed algorithms significantly outperform other well-known systems in terms of recognition rate.

This paper proposes a novel nonlinear discriminant analysis method named by kernerlized maximum average margin criterion (KMAMC), which has combined the idea of support vector machine with the kernel fisher discriminant analysis (KFD). We also use a simple method to prove the relationship between both kernel methods. The difference of KMAMC from traditional KFD methods include: (1) the within-class and between-class scatter matrices are computed based on the support vectors instead of all the samples; (2) multiple centers are exploited instead of the single center in computing the two scatter matrices; (3) the discriminant criteria is formulated as subtracting the trace of within-class scatter matrix from that of the between-class scatter matrix, therefore, the tedious singularity problem is avoided. These features have made KMAMC more practical for real-world applications. Our experiments on two face databases, the FERET and CAS-PEAL face database, have illustrated its excellent performance compared with some traditional methods such as Eigenface, Fisherface, and KFD.