Dataset Condensation (DC) aims to obtain a condensed dataset that allows models trained on the condensed dataset to achieve performance comparable to those trained on the full dataset. Recent DC approaches increasingly focus on encoding knowledge into realistic images with soft labeling, for their scalability to ImageNet-scale datasets and strong capability of cross-domain generalization. However, this strong performance comes at a substantial storage cost which could significantly exceed the storage cost of the original dataset. We argue that the three key properties to alleviate this performance-storage dilemma are informativeness, discriminativeness, and compressibility of the condensed data. Towards this end, this paper proposes a Soft label compression-centric dataset condensation framework using COding RatE (SCORE). SCORE formulates dataset condensation as a min-max optimization problem, which aims to balance the three key properties from an information-theoretic perspective. In particular, we theoretically demonstrate that our coding rate-inspired objective function is sub-modular, and its optimization naturally enforces low-rank structure in the soft label set corresponding to each condensed data. Extensive experiments on large-scale datasets, including ImageNet-1K and Tiny-ImageNet, demonstrate that SCORE outperforms existing methods in most cases. Even with a 30× compression of soft labels, performance decreases by only 5.5% and 2.7% for ImageNet-1K with IPC 10 and 50, respectively. The code is available at https://github.com/KeViNYuAn0314/SCORE.

• Semantic-aware diffusion: edits guided by image-text discrepancy for precise, controllable changes. • Fast and lightweight: 50-iteration fine-tune with LoRA ( 0.37% params), 17 s per image. • Better alignment without sacrificing fidelity: higher CLIP on TEdBench with competitive LPIPS. • Broadly applicable: runs on SD v1.4 and ports to SDXL (via IP-Adapter) with consistent gains. Text-guided single-image editing has emerged as a promising solution to precisely alter an input image based on the target texts, such as making a standing dog appear seated or a bird spreading its wings. While effective, conventional approaches require a two-step process, including fine-tuning the target text embedding for over 1K iterations and the generative model for another 1.5K iterations. Although it ensures that the resulting image closely aligns with both the input image and the target text, this process often requires 7 minutes per image, posing a challenge for practical application due to its time-intensive nature. To address this bottleneck, we introduce FastEdit, a fast text-guided single-image editing method with semantic-aware diffusion fine-tuning, accelerating the editing process to just 17 seconds. FastEdit streamlines the generative model’s fine-tuning phase, reducing it from 1.5K to 50 iterations. Specifically, we perform diffusion fine-tuning on certain time steps determined by the semantic discrepancy between the input image and target text. We conduct extensive experiments to validate the editing performance of our approach and show promising editing capabilities, including content addition, style transfer, background replacement, and posture manipulation, etc. Our code and more edited images are available at https://fastedit-sd.github.io .

Recently, significant progress has been made in text-video retrieval by adapting CLIP to the text-video domain. To mitigate the computational overhead of full fine-tuning, current studies have shifted focus toward parameter-efficient fine-tuning strategies, such as Adapters, Prompts, and LoRA. However, most methods generally obtain the global video representation by frame-level aggregation,i.e.,pooling operation over [CLS] token in each frame, which fails to capture the local details. Meanwhile, these methods process all visual tokens indiscriminately when learning video representations, overlooking the interference introduced by redundant tokens. To tackle these issues, this paper proposes aDual-Attention Video Representation Learning(DAVRL) framework which synergistically combines dual-attention based global-local interactions with token selection. Specifically, we incorporate a dual-attention global-local interaction paradigm and formulate a progressive cross-frame interaction module, which implement global interactions between well-designed learnable global video tokens and all visual tokens, and progressive local interactions from single-frame to multi-frame scales. Moreover, we propose a semantic-aware token selection module to prune the spatial-temporal redundant tokens, under the guidance of both global and local semantics. As such, our DAVRL framework can learn more informative yet discriminative global video representations while mitigating the interference of redundant tokens. Extensive experiments demonstrate the superiority of our DAVRL on diverse datasets with only0.32% tunable parameters.

End-to-end autonomous driving frameworks enable seamless integration of perception and planning but often rely on one-shot trajectory prediction, which may lead to unstable control and vulnerability to occlusions in single-frame perception. To address this, we propose the Momentum-Aware Driving (MomAD) framework, which introduces trajectory momentum and perception momentum to stabilize and refine trajectory predictions. MomAD comprises two core components: (1) Topological Trajectory Matching (TTM) employs Hausdorff Distance to select the optimal planning query that aligns with prior paths to ensure coherence; (2) Momentum Planning Interactor (MPI) cross-attends the selected planning query with historical queries to expand static and dynamic perception files. This enriched query, in turn, helps regenerate long-horizon trajectory and reduce collision risks. To mitigate noise arising from dynamic environments and detection errors, we introduce robust instance denoising during training, enabling the planning model to focus on critical signals and improve its robustness. We also propose a novel Trajectory Prediction Consistency (TPC) metric to quantitatively assess planning stability. Experiments on the nuScenes dataset demonstrate that MomAD achieves superior long-term consistency (≥ 3s) compared to SOTA methods. Moreover, evaluations on the curated Turning-nuScenes shows that MomAD reduces the collision rate by 26% and improves TPC by 0.97m (33.45%) over a 6s prediction horizon, while closed- loop on Bench2Drive demonstrates an up to 16.3% improvement in success rate. The source code is available at https://github.com/adept-thu/MomAD.

LiDAR-based 3D object detection is crucial for various applications but often experiences performance degradation in real-world deployments due to domain shifts. While most studies focus on cross-dataset shifts, such as changes in environments and object geometries, practical corruptions from sensor variations and weather conditions remain underexplored. In this work, we propose a novel online test-time adaptation framework for 3D detectors that effectively tackles these shifts, including a challenging $\textit{cross-corruption}$ scenario where cross-dataset shifts and corruptions co-occur. By leveraging long-term knowledge from previous test batches, our approach mitigates catastrophic forgetting and adapts effectively to diverse shifts. Specifically, we propose a Model Synergy (MOS) strategy that dynamically selects historical checkpoints with diverse knowledge and assembles them to best accommodate the current test batch. This assembly is directed by our proposed Synergy Weights (SW), which perform a weighted averaging of the selected checkpoints, minimizing redundancy in the composite model. The SWs are computed by evaluating the similarity of predicted bounding boxes on the test data and the independence of features between checkpoint pairs in the model bank. To maintain an efficient and informative model bank, we discard checkpoints with the lowest average SW scores, replacing them with newly updated models. Our method was rigorously tested against existing test-time adaptation strategies across three datasets and eight types of corruptions, demonstrating superior adaptability to dynamic scenes and conditions. Notably, it achieved a 67.3% improvement in a challenging cross-corruption scenario, offering a more comprehensive benchmark for adaptation. Source code: https://github.com/zhuoxiao-chen/MOS.

Linear attention has emerged as a promising alternative to softmax-based attention, leveraging kernelized feature maps to reduce complexity from quadratic to linear in sequence length. However, the non-negative constraint on feature maps and the relaxed exponential function used in approximation lead to significant information loss compared to the original query-key dot products, resulting in less discriminative attention maps with higher entropy. To address the missing interactions driven by negative values in query-key pairs, we propose a polarity-aware linear attention mechanism that explicitly models both same-signed and opposite-signed query-key interactions, ensuring comprehensive coverage of relational information. Furthermore, to restore the spiky properties of attention maps, we provide a theoretical analysis proving the existence of a class of element-wise functions (with positive first and second derivatives) that can reduce entropy in the attention distribution. For simplicity, and recognizing the distinct contributions of each dimension, we employ a learnable power function for rescaling, allowing strong and weak attention signals to be effectively separated. Extensive experiments demonstrate that the proposed PolaFormer improves performance on various vision tasks, enhancing both expressiveness and efficiency by up to 4.6%.

Wheat management strategies play a critical role in determining yield. Traditional management decisions often rely on labour-intensive expert inspections, which are expensive, subjective and difficult to scale. Recently, Vision-Language Models (VLMs) have emerged as a promising solution to enable scalable, data-driven management support. However, due to a lack of domain-specific knowledge, directly applying VLMs to wheat management tasks results in poor quantification and reasoning capabilities, ultimately producing vague or even misleading management recommendations. In response, we propose WisWheat, a wheat-specific dataset with a three-layered design to enhance VLM performance on wheat management tasks: (1) a foundational pretraining dataset of 47,871 image-caption pairs for coarsely adapting VLMs to wheat morphology; (2) a quantitative dataset comprising 7,263 VQA-style image-question-answer triplets for quantitative trait measuring tasks; and (3) an Instruction Fine-tuning dataset with 4,888 samples targeting biotic and abiotic stress diagnosis and management plan for different phenological stages. Extensive experimental results demonstrate that fine-tuning open-source VLMs (e.g., Qwen2.5 7B) on our dataset leads to significant performance improvements. Specifically, the Qwen2.5 VL 7B fine-tuned on our wheat instruction dataset achieves accuracy scores of 79.2% and 84.6% on wheat stress and growth stage conversation tasks respectively, surpassing even general-purpose commercial models such as GPT-4o by a margin of 11.9% and 34.6%.

Text-to-video (T2V) synthesis has advanced rapidly, yet current evaluation metrics primarily capture visual quality and temporal consistency, offering limited insight into how synthetic videos perform in downstream tasks such as text-to-video retrieval (TVR). In this work, we introduce SynTVA, a new dataset and benchmark designed to evaluate the utility of synthetic videos for building retrieval models. Based on 800 diverse user queries derived from MSRVTT training split, we generate synthetic videos using state-of-the-art T2V models and annotate each video-text pair along four key semantic alignment dimensions: Object & Scene, Action, Attribute, and Prompt Fidelity. Our evaluation framework correlates general video quality assessment (VQA) metrics with these alignment scores, and examines their predictive power for downstream TVR performance. To explore pathways of scaling up, we further develop an Auto-Evaluator to estimate alignment quality from existing metrics. Beyond benchmarking, our results show that SynTVA is a valuable asset for dataset augmentation, enabling the selection of high-utility synthetic samples that measurably improve TVR outcomes. Project page and dataset can be found at https://jasoncodemaker.github.io/SynTVA/.

Recent advances in *inverse problem* solving have increasingly adopted flow *priors* over diffusion models due to their ability to construct straight probability paths from noise to data, thereby enhancing efficiency in both training and inference. However, current flow-based inverse solvers face two primary limitations: (i) they operate directly in pixel space, which demands heavy computational resources for training and restricts scalability to high-resolution images, and (ii) they employ guidance strategies with *prior*-agnostic posterior covariances, which can weaken alignment with the generative trajectory and degrade posterior coverage. In this paper, we propose **LFlow** (**L**atent Refinement via **Flow**s), a *training-free* framework for solving linear inverse problems via pretrained latent flow priors. LFlow leverages the efficiency of flow matching to perform ODE sampling in latent space along an optimal path. This latent formulation further allows us to introduce a theoretically grounded posterior covariance, derived from the optimal vector field, enabling effective flow guidance. Experimental results demonstrate that our proposed method outperforms state-of-the-art latent diffusion solvers in reconstruction quality across most tasks.

Maintaining robust 3D perception under dynamic and unpredictable test-time conditions remains a critical challenge for autonomous driving systems. Existing test-time adaptation (TTA) methods often fail in high-variance tasks like 3D object detection due to unstable optimization and sharp minima. While recent model merging strategies based on linear mode connectivity (LMC) offer improved stability by interpolating between fine-tuned checkpoints, they are computationally expensive, requiring repeated checkpoint access and multiple forward passes. In this paper, we introduce CodeMerge, a lightweight and scalable model merging framework that bypasses these limitations by operating in a compact latent space. Instead of loading full models, CodeMerge represents each checkpoint with a low-dimensional fingerprint derived from the source model’s penultimate features and constructs a key-value codebook. We compute merging coefficients using ridge leverage scores on these fingerprints, enabling efficient model composition without compromising adaptation quality. Our method achieves strong performance across challenging benchmarks, improving end-to-end 3D detection 14.9\% NDS on nuScenes-C and LiDAR-based detection by over 7.6\% mAP on nuScenes-to-KITTI, while benefiting downstream tasks such as online mapping, motion prediction and planning even without training. The code is released at \url{https://github.com/UQHTy/CodeMerge}.

Pre-training vision-language representations on human action videos has emerged as a promising approach to reduce reliance on large-scale expert demonstrations for training embodied agents. However, prior methods often employ time con- trastive learning based on goal-reaching heuristics, progressively aligning language instructions from the initial to the final frame. This overemphasis on future frames can result in erroneous vision-language associations, as actions may terminate early or include irrelevant moments in the end. To address this issue, we propose Action Temporal Coherence Learning (AcTOL) to learn ordered and continuous vision-language representations without rigid goal-based constraint. AcTOL treats a video as a continuous trajectory where it (1) contrasts semantic differences be- tween frames to reflect their natural ordering, and (2) imposes a local Brownian bridge constraint to ensure smooth transitions across intermediate frames. Exten- sive imitation learning experiments on both simulated and real robots show that the pretrained features significantly enhance downstream manipulation tasks with high robustness to different linguistic styles of instructions, offering a viable pathway toward generalized embodied agents. Our project page is at https://actol-pretrain.github.io/.

Abstract Detecting out-of-distribution (OOD) inputs plays a crucial role in guaranteeing the reliability of deep neural networks (DNNs) when deployed in real-world scenarios. However, DNNs typically exhibit overconfidence in OOD samples, which is attributed to the similarity in patterns between OOD and in-distribution (ID) samples. To mitigate this overconfidence, advanced approaches suggest the incorporation of auxiliary OOD samples during model training, where the outliers are assigned with an equal likelihood of belonging to any category. However, identifying outliers that share patterns with ID samples poses a significant challenge. To address the challenge, we propose a novel method, V irtual O utlier S m o othing (VOSo), which constructs auxiliary outliers using ID samples, thereby eliminating the need to search for OOD samples. Specifically, VOSo creates these virtual outliers by perturbing the semantic regions of ID samples and infusing patterns from other ID samples. For instance, a virtual outlier might consist of a cat’s face with a dog’s nose, where the cat’s face serves as the semantic feature for model prediction. Meanwhile, VOSo adjusts the labels of virtual OOD samples based on the extent of semantic region perturbation, aligning with the notion that virtual outliers may contain ID patterns. Extensive experiments are conducted on diverse OOD detection benchmarks, demonstrating the effectiveness of the proposed VOSo. Our code will be available at https://github.com/junz-debug/VOSo .

Abstract This article presents a comprehensive survey of online test-time adaptation (OTTA), focusing on effectively adapting machine learning models to distributionally different target data upon batch arrival. Despite the recent proliferation of OTTA methods, conclusions from previous studies are inconsistent due to ambiguous settings, outdated backbones, and inconsistent hyperparameter tuning, which obscure core challenges and hinder reproducibility. To enhance clarity and enable rigorous comparison, we classify OTTA techniques into three primary categories and benchmark them using a modern backbone, the Vision Transformer. Our benchmarks cover conventional corrupted datasets such as CIFAR-10/100-C and ImageNet-C, as well as real-world shifts represented by CIFAR-10.1, OfficeHome, and CIFAR-10-Warehouse. The CIFAR-10-Warehouse dataset includes a variety of variations from different search engines and synthesized data generated through diffusion models. To measure efficiency in online scenarios, we introduce novel evaluation metrics, including GFLOPs, wall clock time, and GPU memory usage, providing a clearer picture of the trade-offs between adaptation accuracy and computational overhead. Our findings diverge from existing literature, revealing that (1) transformers demonstrate heightened resilience to diverse domain shifts, (2) the efficacy of many OTTA methods relies on large batch sizes, and (3) stability in optimization and resistance to perturbations are crucial during adaptation, particularly when the batch size is 1. Based on these insights, we highlight promising directions for future research. Our benchmarking toolkit and source code are available at https://github.com/Jo-wang/OTTA_ViT_survey .

LiDAR-based 3D object detection has recently seen significant advancements through active learning (AL), attaining satisfactory performance by training on a small fraction of strategically selected point clouds. However, in real-world deployments where streaming point clouds may include unknown or novel objects, the ability of current AL methods to capture such objects remains unexplored. This paper investigates a more practical and challenging research task: Open World Active Learning for 3D Object Detection (OWAL-3D), aimed at acquiring informative point clouds with new concepts. To tackle this challenge, we propose a simple yet effective strategy called Open Label Conciseness (OLC), which mines novel 3D objects with minimal annotation costs. Our empirical results show that OLC successfully adapts the 3D detection model to the open world scenario with just a single round of selection. Any generic AL policy can then be integrated with the proposed OLC to efficiently address the OWAL-3D problem. Based on this, we introduce the Open-CRB framework, which seamlessly integrates OLC with our preliminary AL method, CRB, designed specifically for 3D object detection. We develop a comprehensive codebase for easy reproducing and future research, supporting 15 baseline methods (i.e., active learning, out-of-distribution detection and open world detection), 2 types of modern 3D detectors (i.e., one-stage SECOND and two-stage PV-RCNN) and 3 benchmark 3D datasets (i.e., KITTI, nuScenes and Waymo). Extensive experiments evidence that the proposed Open-CRB demonstrates superiority and flexibility in recognizing both novel and known classes with very limited labeling costs, compared to state-of-the-art baselines.

Post-hoc out-of-distribution (OOD) detection has garnered intensive attention in reliable machine learning. Many efforts have been dedicated to deriving score functions based on logits, distances, or rigorous data distribution assumptions to identify low-scoring OOD samples. Nevertheless, these estimate scores may fail to accurately reflect the true data density or impose impractical constraints. To provide a unified perspective on density-based score design, we propose a novel theoretical framework grounded in Bregman divergence, which extends distribution considerations to encompass an exponential family of distributions. Leveraging the conjugation constraint revealed in our theorem, we introduce a \textsc{ConjNorm} method, reframing density function design as a search for the optimal norm coefficient $p$ against the given dataset. In light of the computational challenges of normalization, we devise an unbiased and analytically tractable estimator of the partition function using the Monte Carlo-based importance sampling technique. Extensive experiments across OOD detection benchmarks empirically demonstrate that our proposed \textsc{ConjNorm} has established a new state-of-the-art in a variety of OOD detection setups, outperforming the current best method by up to 13.25\% and 28.19\% (FPR95) on CIFAR-100 and ImageNet-1K, respectively.

Collecting and curating data plays a crucial role in environmental surveying. In order to gather meaningful samples, it is often necessary to develop a real-time data curation system that processes data on-the-fly and enriches it with validated models from domain experts. One area where this is particularly important is underwater marine surveys, where the vastness of the sea requires human interaction in the curation process to focus on relevant areas for exploration. Additionally, ongoing surveys are susceptible to poor performance due to data drift, which hinders the ability to provide valuable feedback for guiding data collection. While recent advancements have shown promise in addressing these challenges, they often overlook the practical constraints associated with remote data collection, such as limited processing power and latency. To overcome these limitations, this paper proposes a real-time system that adapts to data drift and enables the recording of uncertain samples for further processing and analysis on shore. The results of our approach demonstrate a remarkable improvement in species recognition, achieving an almost 18% improvement compared to the best baseline method in unseen areas. Importantly, this improvement is achieved while meeting the real-time requirements of surveys and consuming only 15W of power. By effectively addressing the challenges of underwater data drift, our proposed approach provides an efficient and effective solution for environmental surveys.

LiDAR-based 3D object detection has seen impressive advances in recent times. However, deploying trained 3D detectors in the real world often yields unsatisfactory performance when the distribution of the test data significantly deviates from the training data due to different weather conditions, object sizes, etc. A key factor in this performance degradation is the diminished generalizability of pre-trained models, which creates a sharp loss landscape during training. Such sharpness, when encountered during testing, can precipitate significant performance declines, even with minor data variations. To address the aforementioned challenges, we propose dual-perturbation optimization (DPO) for Test-time Adaptation in 3D Object Detection (TTA-3OD). We minimize the sharpness to cultivate a flat loss landscape to ensure model resiliency to minor data variations, thereby enhancing the generalization of the adaptation process. To fully capture the inherent variability of the test point clouds, we further introduce adversarial perturbation to the input BEV features to better simulate the noisy test environment. As the dual perturbation strategy relies on trustworthy supervision signals, we utilize a reliable Hungarian matcher to filter out pseudo-labels sensitive to perturbations. Additionally, we introduce early Hungarian cutoff to avoid error accumulation from incorrect pseudo-labels by halting the adaptation process. Extensive experiments across three types of transfer tasks demonstrate that the proposed DPO significantly surpasses previous state-of-the-art approaches, specifically on Waymo → KITTI, outperforming the most competitive baseline by 57.72% in AP3D and reaching 91% of the fully supervised upper bound. Our code is available at https://github.com/Jo-wang/DPO.

Class-agnostic object detection (OD) can be a cornerstone or a bottleneck for many downstream vision tasks. Despite considerable advancements in bottom-up and multi-object discovery methods that leverage basic visual cues to identify salient objects, consistently achieving a high recall rate remains difficult due to the diversity of object types and their contextual complexity. In this work, we investigate using vision-language models (VLMs) to enhance object detection via a self-supervised prompt learning strategy. Our initial findings indicate that manually crafted text queries often result in undetected objects, primarily because detection confidence diminishes when the query words exhibit semantic overlap. To address this, we propose a Dispersing Prompt Expansion (DiPEx) approach. DiPEx progressively learns to expand a set of distinct, non-overlapping hyperspherical prompts to enhance recall rates, thereby improving performance in downstream tasks such as out-of-distribution OD. Specifically, DiPEx initiates the process by self-training generic parent prompts and selecting the one with the highest semantic uncertainty for further expansion. The resulting child prompts are expected to inherit semantics from their parent prompts while capturing more fine-grained semantics. We apply dispersion losses to ensure high inter-class discrepancy among child prompts while preserving semantic consistency between parent-child prompt pairs. To prevent excessive growth of the prompt sets, we utilize the maximum angular coverage (MAC) of the semantic space as a criterion for early termination. We demonstrate the effectiveness of DiPEx through extensive class-agnostic OD and OOD-OD experiments on MS-COCO and LVIS, surpassing other prompting methods by up to 20.1% in AR and achieving a 21.3% AP improvement over SAM.

In this paper, we tackle the challenge of generating high-quality hash codes for cross-modal retrieval in the presence of incomplete labels, which creates uncertainty in distinguishing between positive and negative pairs. Vision-language models such as CLIP offer a potential solution by providing generic knowledge for missing label recovery, yet their zero-shot performance remains insufficient. To address this, we propose a novel Prompt Contrastive Recovery approach, \textbf{PCRIL}, which progressively identifies promising positive classes from unknown label sets and recursively searches for other relevant labels. Identifying unknowns is nontrivial due to the fixed and long-tailed patterns of positive label sets in training data, which hampers the discovery of new label combinations. Therefore, we consider each subset of positive labels and construct three types of negative prompts through deletion, addition, and replacement for prompt learning. The augmented supervision guides the model to measure the completeness of label sets, thus facilitating the subsequent greedy tree search for label completion. We also address extreme cases of significant unknown labels and lack of negative pairwise supervision by deriving two augmentation strategies: seeking unknown-complementary samples for mixup and random flipping for negative labels. Extensive experiments reveal the vulnerability of current methods and demonstrate the effectiveness of PCRIL, achieving an average 12\% mAP improvement to the current SOTA across all datasets. Our code is available at https://github.com/E-Galois/PCRIL.

Dataset Distillation (DD) is designed to generate condensed representations of extensive image datasets, enhancing training efficiency. Despite recent advances, there remains considerable potential for improvement, particularly in addressing the notable redundancy within the color space of distilled images. In this paper, we propose a two-fold optimization strategy to minimize color redundancy at the individual image and overall dataset levels, respectively. At the image level, we employ a palette network, a specialized neural network, to dynamically allocate colors from a reduced color space to each pixel. The palette network identifies essential areas in synthetic images for model training, and consequently assigns more unique colors to them. At the dataset level, we develop a color-guided initialization strategy to minimize redundancy among images. Representative images with the least replicated color patterns are selected based on the information gain. A comprehensive performance study involving various datasets and evaluation scenarios is conducted, demonstrating the superior performance of our proposed color-aware DD compared to existing DD methods.

In the realm of modern autonomous driving, the perception system is indispensable for accurately assessing the state of the surrounding environment, thereby enabling informed prediction and planning. The key step to this system is related to 3D object detection that utilizes vehicle-mounted sensors such as LiDAR and cameras to identify the size, the category, and the location of nearby objects. Despite the surge in 3D object detection methods aimed at enhancing detection precision and efficiency, there is a gap in the literature that systematically examines their resilience against environmental variations, noise, and weather changes. This study emphasizes the importance of robustness, alongside accuracy and latency, in evaluating perception systems under practical scenarios. Our work presents an extensive survey of camera-only, LiDAR-only, and multi-modal 3D object detection algorithms, thoroughly evaluating their trade-off between accuracy, latency, and robustness, particularly on datasets like KITTI-C and nuScenes-C to ensure fair comparisons. Among these, multi-modal 3D detection approaches exhibit superior robustness, and a novel taxonomy is introduced to reorganize the literature for enhanced clarity. This survey aims to offer a more practical perspective on the current capabilities and the constraints of 3D object detection algorithms in real-world applications, thus steering future research towards robustness-centric advancements.

Unsupervised domain adaptation (DA) with the aid of pseudo labeling techniques has emerged as a crucial approach for domain-adaptive 3D object detection. While effective, existing DA methods suffer from a substantial drop in performance when applied to a multi-class training setting, due to the co-existence of low-quality pseudo labels and class imbalance issues. In this paper, we address this challenge by proposing a novel ReDB framework tailored for learning to detect all classes at once. Our approach produces Reliable, Diverse, and class-Balanced pseudo 3D boxes to iteratively guide the self-training on a distributionally different target domain. To alleviate disruptions caused by the environmental discrepancy (e.g., beam numbers), the proposed cross-domain examination (CDE) assesses the correctness of pseudo labels by copy-pasting target instances into a source environment and measuring the prediction consistency. To reduce computational overhead and mitigate the object shift (e.g., scales and point densities), we design an overlapped boxes counting (OBC) metric that allows to uniformly downsample pseudo-labeled objects across different geometric characteristics. To confront the issue of inter-class imbalance, we progressively augment the target point clouds with a class-balanced set of pseudo-labeled target instances and source objects, which boosts recognition accuracies on both frequently appearing and rare classes. Experimental results on three benchmark datasets using both voxel-based (i.e., SECOND) and point-based 3D detectors (i.e., PointRCNN) demonstrate that our proposed ReDB approach outperforms existing 3D domain adaptation methods by a large margin, improving 23.15% mAP on the nuScenes → KITTI task. The code is available at https://github.com/zhuoxiao-chen/ReDB-DA-3Ddet.

Achieving a reliable LiDAR-based object detector in autonomous driving is paramount, but its success hinges on obtaining large amounts of precise 3D annotations. Active learning (AL) seeks to mitigate the annotation burden through algorithms that use fewer labels and can attain performance comparable to fully supervised learning. Although AL has shown promise, current approaches prioritize the selection of unlabeled point clouds with high uncertainty and/or diversity, leading to the selection of more instances for labeling and reduced computational efficiency. In this paper, we resort to a novel kernel coding rate maximization (Kecor) strategy which aims to identify the most informative point clouds to acquire labels through the lens of information theory. Greedy search is applied to seek desired point clouds that can maximize the minimal number of bits required to encode the latent features. To determine the uniqueness and informativeness of the selected samples from the model perspective, we construct a proxy network of the 3D detector head and compute the outer product of Jacobians from all proxy layers to form the empirical neural tangent kernel (NTK) matrix. To accommodate both one-stage (i.e., Second) and two-stage detectors (i.e., Pv-rcnn), we further incorporate the classification entropy maximization and well trade-off between detection performance and the total number of bounding boxes selected for annotation. Extensive experiments conducted on two 3D benchmarks and a 2D detection dataset evidence the superiority and versatility of the proposed approach. Our results show that approximately 44% box-level annotation costs and 26% computational time are reduced compared to the state-of-the-art AL method, without compromising detection performance. Source code: https://github.com/Luoyadan/KECOR-active-3Ddet.

This paper focuses on model transferability estimation, i.e., assessing the performance of pre-trained models on a downstream task without performing fine-tuning. Motivated by neural collapse (NC) [25] that reveals particular feature geometry at the terminal stage of training, we consider model transferability as how far the target activations obtained by pre-trained models are from their hypothetical state in the terminal phase of the model fine-tuned on the target domain. We propose a metric that measures this proximity based on three phenomena of NC: within-class variability collapse, simplex encoded label interpolation geometry structure is formed, and the nearest center classifier becomes optimal on training data. Through experiments on 11 datasets, we confirm none of the three NC proxies are dispensable, which allows us to obtain very competitive transferability estimation accuracy with approximately 10× wall-clock time speed up compared to state-of-the-art approaches.

Existing Building Damage Detection (BDD) methods always require labour-intensive pixel-level annotations of buildings and their conditions, hence largely limiting their applications. In this paper, we investigate a challenging yet practical scenario of BDD, Unsupervised Building Damage Detection (U-BDD), where only unlabelled pre- and post-disaster satellite image pairs are provided. As a pilot study, we have first proposed an advanced U-BDD baseline that leverages pre-trained vision-language foundation models to address the U-BDD task. However, the apparent domain gap between satellite and generic images causes low confidence in the foundation models used to identify buildings and their damages. In response, we further present a novel self-supervised framework, U-BDD++, which improves upon the U-BDD baseline by addressing domain-specific issues associated with satellite imagery. Extensive experiments on the widely used building damage assessment benchmark demonstrate the effectiveness of the proposed method for unsupervised building damage detection. The presented annotation-free and foundation model-based paradigm ensures an efficient learning phase. This study opens a new direction for real-world BDD and sets a strong baseline for future research.

Road surveying plays a vital role in effective road network management for local governments. However, current practices pose challenges due to their costly, time-consuming, and inaccurate nature. In this paper, we propose an automated survey platform that supports weed, defect and asset monitoring with instance segmentation models. Empowered by recent advancements in vision-language models (VLMs), our solution offers improved flexibility for novel tasks with a limited label set. For domain specific classes, such as pavement cracks and potholes, we train a detector to identify their location given our sparsely annotated images, and alleviate false-positives by rejecting predictions outside regions of interest identified by VLMs. The proposed system directly involves managers in the survey process through a mobile application. The application allows three core functions: 1) capture and cloud upload, 2) real-time survey trajectory monitoring, 3) open-vocabulary detection.

The prevalence of domain adaptive semantic segmentation has prompted concerns regarding source domain data leakage, where private information from the source domain could inadvertently be exposed in the target domain. To circumvent the requirement for source data, source-free domain adaptation has emerged as a viable solution that leverages self-training methods to pseudo-label high-confidence regions and adapt the model to the target data. However, the confidence scores obtained are often highly biased due to overconfidence and class-imbalance issues, which render both model selection and optimization problematic. In this paper, we propose a novel calibration-guided source-free domain adaptive semantic segmentation (Cal-SFDA) framework. The core idea is to estimate the expected calibration error (ECE) from the segmentation predictions, serving as a strong indicator of the model's generalization capability to the unlabeled target domain. The estimated ECE scores, in turn, assist the model training and fair selection in both source training and target adaptation stages. During model pre-training on the source domain, we ensure the differentiability of the ECE objective by leveraging the LogSumExp trick and using ECE scores to select the best source checkpoints for adaptation. To enable ECE estimation on the target domain without requiring labels, we train a value net for ECE estimation and apply statistic warm-up on its BatchNorm layers for stability. The estimated ECE scores assist in determining the reliability of prediction and enable class-balanced pseudo-labeling by positively guiding the adaptation progress and inhibiting potential error accumulation. Extensive experiments on two widely-used synthetic-to-real transfer tasks show that the proposed approach surpasses previous state-of-the-art by up to 5.25% of mIoU with fair model selection criteria.

This work investigates the Single Domain Generalization (SDG) problem and aims to generalize a model from a single source (i.e., training) domain to multiple target (i.e., test) domains coming from different distributions. Most of the existing SDG approaches focus on generating out-of-domain samples by either transforming the source images into different styles or optimizing adversarial noise perturbations applied on the source images. In this paper, we show that generating images with diverse styles can be complementary to creating hard samples when handling the SDG task, and propose our approach of Factorized Frequency Modification (FFM) to fulfill this requirement. Specifically, we design a unified framework consisting of a style transformation module, an adversarial perturbation module, and a dynamic frequency selection module. We seamlessly equip the framework with iterative adversarial training that facilitates learning discriminative features from hard and diverse augmented samples. Extensive experiments are performed on four image recognition benchmark datasets of Digits, CIFAR-10-C, CIFAR-100-C, and PACS, which demonstrates that our method outperforms existing state-of-the-art approaches.

Open-set domain adaptation (OSDA) aims to transfer knowledge from a label-rich source domain to a label-scarce target domain while addressing disturbances from irrelevant target classes not present in the source data. However, most OSDA approaches are limited due to the lack of essential theoretical analysis of generalization bound, reliance on the coexistence of source and target data during adaptation, and failure to accurately estimate model predictions' uncertainty. To address these limitations, the Progressive Graph Learning (PGL) framework is proposed. PGL decomposes the target hypothesis space into shared and unknown subspaces and progressively pseudo-labels the most confident known samples from the target domain for hypothesis adaptation. PGL guarantees a tight upper bound of the target error by integrating a graph neural network with episodic training and leveraging adversarial learning to close the gap between the source and target distributions. The proposed approach also tackles a more realistic source-free open-set domain adaptation (SF-OSDA) setting that makes no assumptions about the coexistence of source and target domains. In a two-stage framework, the SF-PGL model' uniformly selects the most confident target instances from each category at a fixed ratio, and the confidence thresholds in each class weigh the classification loss in the adaptation step. The proposed methods are evaluated on benchmark image classification and action recognition datasets, where they demonstrate superiority and flexibility in recognizing both shared and unknown categories. Additionally, balanced pseudo-labeling plays a significant role in improving calibration, making the trained model less prone to over- or under-confident predictions on the target data.

Signed link prediction in social networks aims to reveal the underlying relationships (i.e., links) among users (i.e., nodes) given their existing positive and negative interactions observed. Most of the prior efforts are devoted to learning node embeddings with graph neural networks (GNNs), which preserve the signed network topology by message-passing along edges to facilitate the downstream link prediction task. Nevertheless, the existing graph-based approaches could hardly provide human-intelligible explanations for the following three questions: (1) which neighbors to aggregate, (2) which path to propagate along, and (3) which social theory to follow in the learning process. To answer the aforementioned questions, in this paper, we investigate how to reconcile thebalanceandstatussocial rules with information theory and develop a unified framework, termed as Signed Infomax Hyperbolic Graph (SIHG). By maximizing the mutual information between edge polarities and node embeddings, one can identify the most representative neighboring nodes that support the inference of edge sign. Different from existing GNNs that could only group features of friends in the subspace, the proposed SIHG incorporates the signed attention module, which is also capable of pushing hostile users far away from each other to preserve the geometry of antagonism. The polarity of the learned edge attention maps, in turn, provides interpretations of the social theories used in each aggregation. In order to model high-order user relations and complex hierarchies, the node embeddings are projected and measured in a hyperbolic space with a lower distortion. Extensive experiments on four signed network benchmarks demonstrate that the proposed SIHG framework significantly outperforms the state-of-the-arts in signed link prediction.

Generalized Zero-Shot Learning (GZSL) aims to recognize images not only for seen classes but also for unseen ones by transferring semantic-visual relationships from the seen to the unseen classes. It is an intuitive solution to take the advantage of generative models to hallucinate realistic unseen samples based on the knowledge learned from the seen classes. However, due to the generation shifts, the synthesized samples by most existing methods may drift from the real distribution of the unseen data. To address this issue, we propose a novel flow-based generative framework that consists of multiple conditional affine coupling layers for learning unseen data generation. Specifically, we investigate and address three essential problems that trigger the generation shifts,i.e.,semantic inconsistency,variance collapse, andstructure disorder. First, to improve the reflection of the semantic information in the generated samples, we proactively embed the semantic information into the transformation in each conditional affine coupling layer. Second, to promote the intrinsic feature variance of the unseen classes, we introduce a boundary sample mining strategy with entropy maximization to discover ambiguous visual variants of semantic prototypes and hereby calibrate the decision boundary of the classifiers. Third, a relative positioning strategy is proposed to revise the attribute embeddings, guiding which to fully preserve the inter-class geometric structure and further avoid structure disorder in the semantic space. Extensive experimental results on four GZSL benchmark datasets demonstrate that GSMFlow achieves the state-of-the-art performance on GZSL.

A typical multi-source domain adaptation (MSDA) approach aims to transfer knowledge learned from a set of labeled source domains, to an unlabeled target domain. Neverthe-less, prior works strictly assume that each source domain shares the identical group of classes with the target domain, which could hardly be guaranteed as the target label space is not observable. In this paper, we consider a more versatile setting of MSDA, namely Generalized Multi-source Domain Adaptation, wherein the source domains are partially over-lapped, and the target domain is allowed to contain novel cat-egories that are not presented in any source domains. This new setting is more elusive than any existing domain adaptation protocols due to the coexistence of the domain and cate-gory shifts across the source and target domains. To address this issue, we propose a variational domain disentanglement (VDD) framework, which decomposes the domain represen-tations and semantic features for each instance by encouraging dimension-wise independence. To identify the target sam-ples of unknown classes, we leverage online pseudo labeling, which assigns the pseudo-labels to unlabeled target data based on the confidence scores. Quantitative and qualitative experi-ments conducted on two benchmark datasets demonstrate the validity of the proposed framework.

The difficulty of image-text retrieval is further exacerbated by the phenomenon of one-to-many correspondence, where multiple semantic manifestations of the other modality could be obtained by a given query. However, the prevailing methods adopt the deterministic embedding strategy to retrieve the most similar candidate, which encodes the representations of different modalities as single points in vector space. We argue that such a deterministic point mapping is obviously insufficient to represent a potential set of retrieval results for one-to-many correspondence, despite its noticeable progress. As a remedy to this issue, we propose a Point to Rectangle Matching (abbreviated as P2RM) mechanism, which actually is a geometric representation learning method for image-text retrieval. Specifically, our intuitive insight is that the representations of different modalities could be extended to rectangles, then a set of points inside such a rectangle embedding could be semantically related to many candidate correspondences. Thus our P2RM method could essentially address the one-to-many correspondence. Besides, we design a novel semantic similarity measurement method from the perspective of distance for our rectangle embedding. Under the evaluation metric for multiple matches, extensive experiments and ablation studies on two commonly used benchmarks demonstrate our effectiveness and superiority in tackling the multiplicity of image-text retrieval.

Generalized zero-shot learning (GZSL) aims to classify samples under the assumption that some classes are not observable during training. To bridge the gap between the seen and unseen classes, most GZSL methods attempt to associate the visual features of seen classes with attributes or to generate unseen samples directly. Nevertheless, the visual features used in the prior approaches do not necessarily encode semantically related information that the shared attributes refer to, which degrades the model generalization to unseen classes. To address this issue, in this paper, we propose a novel semantics disentangling framework for the generalized zero-shot learning task (SDGZSL), where the visual features of unseen classes are firstly estimated by a conditional VAE and then factorized into semantic-consistent and semantic-unrelated latent vectors. In particular, a total correlation penalty is applied to guarantee the independence between the two factorized representations, and the semantic consistency of which is measured by the derived relation network. Extensive experiments conducted on four GZSL benchmark datasets have evidenced that the semantic-consistent features disentangled by the proposed SDGZSL are more generalizable in tasks of canonical and generalized zero-shot learning. Our source code is available at https://github.com/uqzhichen/SDGZSL.

Domain generalization (DG) aims to generalize a model trained on multiple source (i.e., training) domains to a distributionally different target (i.e., test) domain. In contrast to the conventional DG that strictly requires the availability of multiple source domains, this paper considers a more realistic yet challenging scenario, namely Single Domain Generalization (Single-DG), where only one source domain is available for training. In this scenario, the limited diversity may jeopardize the model generalization on unseen target domains. To tackle this problem, we propose a style-complement module to enhance the generalization power of the model by synthesizing images from diverse distributions that are complementary to the source ones. More specifically, we adopt a tractable upper bound of mutual information (MI) between the generated and source samples and perform a two-step optimization iteratively: (1) by minimizing the MI upper bound approximation for each sample pair, the generated images are forced to be diversified from the source samples; (2) subsequently, we maximize the MI between the samples from the same semantic category, which assists the network to learn discriminative features from diverse-styled images. Extensive experiments on three benchmark datasets demonstrate the superiority of our approach, which surpasses the state-of-the-art single-DG methods by up to 25.14%. The code will be publicly available at https://github.com/BUserName/Learning_to_diversify

Being an indispensable component in location-based social networks, next point-of-interest (POI) recommendation recommends users unexplored POIs based on their recent visiting histories. However, existing work mainly models check-in data as isolated POI sequences, neglecting the crucial collaborative signals from cross-sequence check-in information. Furthermore, the sparse POI-POI transitions restrict the ability of a model to learn effective sequential patterns for recommendation. In this paper, we propose Sequence-to-Graph (Seq2Graph) augmentation for each POI sequence, allowing collaborative signals to be propagated from correlated POIs belonging to other sequences. We then devise a novel Sequence-to-Graph POI Recommender (SGRec), which jointly learns POI embeddings and infers a user's temporal preferences from the graph-augmented POI sequence. To overcome the sparsity of POI-level interactions, we further infuse category-awareness into SGRec with a multi-task learning scheme that captures the denser category-wise transitions. As such, SGRec makes full use of the collaborative signals for learning expressive POI representations, and also comprehensively uncovers multi-level sequential patterns for user preference modelling. Extensive experiments on two real-world datasets demonstrate the superiority of SGRec against state-of-the-art methods in next POI recommendation.

Generalized Zero-Shot Learning (GZSL) is the task of leveraging semantic information to recognize seen and unseen samples, where unseen classes are not observable during training. It is natural to derive generative models and hallucinate training samples for unseen classes based on the knowledge learned from the seen samples. However, most of these models suffer from the generation shifts, where the synthesized samples may drift from the real distribution of unseen data. In this paper, we propose a novel generative flow framework that consists of multiple conditional affine coupling layers for learning unseen data generation. In particular, we identify three potential problems that trigger the generation shifts, i.e., semantic inconsistency, variance collapse, and structure disorder and address them respectively. First, to reinforce the correlations between the generated samples and their corresponding attributes, we explicitly embed the semantic information into the transformations in each coupling layer. Second, to recover the intrinsic variance of the real unseen features, we introduce a visual perturbation strategy to diversify the generated data and hereby help adjust the decision boundary of the classifiers. Third, a relative positioning strategy is proposed to revise the attribute embeddings, guiding them to fully preserve the inter-class geometric structure and further avoid structure disorder in the semantic space. Experimental results demonstrate that GSMFlow achieves the state-of-the-art performance on GZSL.

With the advent of media streaming, video action recognition has become progressively important for various applications, yet at the high expense of requiring large-scale data labelling. To overcome the problem of expensive data labelling, domain adaptation techniques have been proposed, which transfer knowledge from fully labelled data (i.e., source domain) to unlabelled data (i.e., target domain). The majority of video domain adaptation algorithms are proposed for closed-set scenarios in which all the classes are shared among the domains. In this work, we propose an open-set video domain adaptation approach to mitigate the domain discrepancy between the source and target data, allowing the target data to contain additional classes that do not belong to the source domain. Different from previous works, which only focus on improving accuracy for shared classes, we aim to jointly enhance the alignment of the shared classes and recognition of unknown samples. Towards this goal, class-conditional extreme value theory is applied to enhance the unknown recognition. Specifically, the entropy values of target samples are modelled as generalised extreme value distributions, which allows separating unknown samples lying in the tail of the distribution. To alleviate the negative transfer issue, weights computed by the distance from the sample entropy to the threshold are leveraged in adversarial learning in the sense that confident source and target samples are aligned, and unconfident samples are pushed away. The proposed method has been thoroughly evaluated on both small-scale and large-scale cross-domain video datasets and achieved the state-of-the-art performance.

With the rapid development of intelligent detection algorithms based on deep learning, much progress has been made in automatic road defect recognition and road marking parsing. This can effectively address the issue of an expensive and time-consuming process for professional inspectors to review the street manually. Towards this goal, we present RoadAtlas, a novel end-to-end integrated system that can support 1) road defect detection, 2) road marking parsing, 3) a web-based dashboard for presenting and inputting data by users, and 4) a backend containing a well-structured database and developed APIs.

Hashing techniques are in great demand for a wide range of real-world applications such as image retrieval and network compression. Nevertheless, existing approaches could hardly guarantee a satisfactory performance with the extremely low-bit (e.g., 4-bit) hash codes due to the severe information loss and the shrink of the discrete solution space. In this article, we propose a novelCollaborative Learningstrategy that is tailored for generating high-quality low-bit hash codes. The core idea is to jointly distill bit-specific and informative representations for a group of pre-defined code lengths. The learning of short hash codes among the group can benefit from the manifold shared with other long codes, where multiple views from different hash codes provide the supplementary guidance and regularization, making the convergence faster and more stable. To achieve that, an asymmetric hashing framework with two variants of multi-head embedding structures is derived, termed as Multi-head Asymmetric Hashing (MAH), leading to great efficiency of training and querying. Extensive experiments on three benchmark datasets have been conducted to verify the superiority of the proposed MAH, and have shown that the 8-bit hash codes generated by MAH achieve 94.3 percent of the MAP11.Mean Average Precision (MAP)score on the CIFAR-10 dataset, which significantly surpasses the performance of the 48-bit codes by the state-of-the-arts in image retrieval tasks.

While deep hashing has made great progress in large-scale multimedia retrieval, most of the existing approaches under-explore the semantic correlations and neglect the effect of context-aware visual learning. In this paper, we propose a dual-stream learning framework, termed as Deep Collaborative Discrete Hashing (DCDH), which constructs a discriminative common discrete space by collaboratively incorporating the shared and individual semantics deduced from visual features and semantics. Specifically, DCDH generates context-aware representations by employing the outer product of visual embeddings and semantic encodings. To further preserve the original semantics and alleviate the class imbalance problem, we introduce the focal loss to take advantage of frequent and rare concepts. Furthermore, a common binary code space is constructed based on the joint learning of the visual representations, the context-aware representations, and the label distribution calibration. Three losses, i.e., the pairwise similarity loss, the quantization loss, and the balanced classification loss, are collaboratively optimized in the general learning framework of DCDH. Extensive experiments conducted on three large-scale benchmark datasets demonstrate the superiority of the proposed method, yielding the state-of-the-art image retrieval performance.

Semantic-preserving hashing establishes efficient multimedia retrieval by transferring knowledge from original data to hash codes so that the latter can preserve the underlying visual and semantic similarities. However, it becomes a crucial bottleneck: how to effectively bridge the trilateral domain gaps (i.e., the visual, semantic, and hashing spaces) to further improve the retrieval accuracy. In this article, we propose an inductive structure consistent hashing (ISCH) method, which can interactively coordinate the semantic correlations between the visual feature space, the binary class space, and the discrete hashing space. Specifically, an inductive semantic space is formulated by a simple multilayer stacking class-encoder, which transforms the naive class information into flexible semantic embeddings. Meanwhile, we design a semantic dictionary learning model to facilitate the bilateral visual-semantic bridging and guide the class-encoder toward reliable semantics, which could well alleviate the visual-semantic bias problem. In particular, the visual descriptors and respective semantic class representations are regularized with a coinciding alignment module. In order to generate privileged hash codes, we further explore semantic and prototype binary code learning to jointly quantify the semantic and latent visual representations into unified discrete hash codes. Moreover, an efficient optimization algorithm is developed to address the resulting discrete programming problem. Comprehensive experiments conducted on four large-scale data sets, i.e., CIFAR-10, NUSWIDE, ImageNet, and MSCOCO, demonstrate the superiority of our method over the state-of-the-art alternatives against different evaluation protocols.

Meta-learning for few-shot learning allows a machine to leverage previously acquired knowledge as a prior, thus improving the performance on novel tasks with only small amounts of data. However, most mainstream models suffer from catastrophic forgetting and insufficient robustness issues, thereby failing to fully retain or exploit long-term knowledge while being prone to cause severe error accumulation. In this paper, we propose a novel Continual Meta-Learning approach with Bayesian Graph Neural Networks (CML-BGNN) that mathematically formulates meta-learning as continual learning of a sequence of tasks. With each task forming as a graph, the intra- and inter-task correlations can be well preserved via message-passing and history transition. To remedy topological uncertainty from graph initialization, we utilize Bayes by Backprop strategy that approximates the posterior distribution of task-specific parameters with amortized inference networks, which are seamlessly integrated into the end-to-end edge learning. Extensive experiments conducted on the miniImageNet and tieredImageNet datasets demonstrate the effectiveness and efficiency of the proposed method, improving the performance by 42.8% compared with state-of-the-art on the miniImageNet 5-way 1-shot classification task.

Domain shift is a fundamental problem in visual recognition which typically arises when the source and target data follow different distributions. The existing domain adaptation approaches which tackle this problem work in the "closed-set" setting with the assumption that the source and the target data share exactly the same classes of objects. In this paper, we tackle a more realistic problem of the "open-set" domain shift where the target data contains additional classes that were not present in the source data. More specifically, we introduce an end-to-end Progressive Graph Learning (PGL) framework where a graph neural network with episodic training is integrated to suppress underlying conditional shift and adversarial learning is adopted to close the gap between the source and target distributions. Compared to the existing open-set adaptation approaches, our approach guarantees to achieve a tighter upper bound of the target error. Extensive experiments on three standard open-set benchmarks evidence that our approach significantly outperforms the state-of-the-arts in open-set domain adaptation.

Image captioning aims to describe an image with a concise, accurate, and interesting sentence. To build such an automatic neural captioner, the traditional models align the generated words with a number of human-annotated sentences to mimic human-like captions. However, the crowd-sourced annotations inevitably come with data quality issues such as grammatical errors, wrong identification of visual objects and sub-optimal sentence focus. During the model training, existing methods treat all the annotations equally regardless of the data quality. In this work, we explicitly engage human consensus to measure the quality of ground truth captions in advance, and directly encourage the model to learn high quality captions with high priority. Therefore, the proposed consensus-oriented method can accelerate the training process and achieve superior performance with only supervised objective without time-consuming reinforcement learning. The novel consensus loss can be implemented into most of the existing state-of-the-art methods, boosting the BLEU-4 performance by maximum relative 12.47% comparing to the conventional cross-entropy loss. Extensive experiments are conducted on MS-COCO Image Captioning dataset demonstrating the proposed human consensus-oriented training method can significantly improve the training efficiency and model effectiveness.

Domain adaptation techniques, which focus on adapting models between distributionally different domains, are rarely explored in the video recognition area due to the significant spatial and temporal shifts across the source (i.e. training) and target (i.e. test) domains. As such, recent works on visual domain adaptation which leverage adversarial learning to unify the source and target video representations and strengthen the feature transferability are not highly effective on the videos. To overcome this limitation, in this paper, we learn a domain-agnostic video classifier instead of learning domain-invariant representations, and propose an Adversarial Bipartite Graph (ABG) learning framework which directly models the source-target interactions with a network topology of the bipartite graph. Specifically, the source and target frames are sampled as heterogeneous vertexes while the edges connecting two types of nodes measure the affinity among them. Through message-passing, each vertex aggregates the features from its heterogeneous neighbors, forcing the features coming from the same class to be mixed evenly. Explicitly exposing the video classifier to such cross-domain representations at the training and test stages makes our model less biased to the labeled source data, which in-turn results in achieving a better generalization on the target domain. The proposed framework is agnostic to the choices of frame aggregation, and therefore, four different aggregation functions are investigated for capturing appearance and temporal dynamics. To further enhance the model capacity and testify the robustness of the proposed architecture on difficult transfer tasks, we extend our model to work in a semi-supervised setting using an additional video-level bipartite graph. Extensive experiments conducted on four benchmark datasets evidence the effectiveness of the proposed approach over the state-of-the-art methods on the task of video recognition.

Even though the multimedia data is ubiquitous on the web, the scarcity of the annotated data and variety of data modalities hinder their usage by multimedia applications. Heterogeneous domain adaptation (HDA) has therefore arisen to address such limitations by facilitating the knowledge transfer between heterogeneous domains. Existing HDA methods only focus on aligning the cross-domain feature distributions and ignore the importance of maximizing the margin among different classes, which may lead to a sub-optimal classification performance. To tackle this problem, in this paper, we propose the Prototype-Matching Graph Network (PMGN), which gradually explores the domain-invariant class prototype representations. Specifically, we build an end-to-end Graph Prototypical Network, which computes the class prototypes through multiple layers of edge learning, node aggregation, and discrepancy minimization. Our framework utilizes the Swap training strategy to provide adequate supervision for training the edge learning component. Moreover, the proposed PMGN can be equipped with the clustering module that utilises the KL-divergence as a distance metric to reduce the distribution difference between the source and target data. Extensive experiments on three HDA tasks (i.e. object recognition, text-to-image classification, and text categorization) demonstrate the superiority of our approach over the state-of-the-art HDA methods.

Existing methods using generative adversarial approaches for Zero-Shot Learning (ZSL) aim to generate realistic visual features from class semantics by a single generative alignment, which is highly under-constrained. As a result, the previous methods cannot guarantee that the generated visual features can truthfully reflect the corresponding semantics. To address this issue, we propose a novel method named Cycle-consistent Adversarial Networks for Zero-Shot Learning (CANZSL). It encourages a visual feature generator to synthesize realistic visual features from semantics, and then inversely translate back the synthesized visual features to the corresponding semantic space by a semantic feature generator. Furthermore, in this paper a more challenging and practical ZSL problem is considered where the original semantics are from natural language with irrelevant words instead of clean semantics, which are widely used in previous work. Specifically, a multi-modal consistent bidirectional generative adversarial model is trained to handle unseen instances by suppressing noise in the natural language. A forward one-to-many mapping from the class level descriptions to the visual features is coupled with an inverse many-to-one mapping from the visual space to the semantic space. Thus, a multi-modal cycle-consistency loss between the synthesized semantic representations and the ground truth can be learned and leveraged to enforce the generated semantic features to approximate to the real distribution in semantic space. Extensive experiments are conducted to demonstrate that our method consistently outperforms state-of-the-art approaches on natural language-based zero-shot learning tasks.

Visual paragraph generation aims to automatically describe a given image from different perspectives and organize sentences in a coherent way. In this paper, we address three critical challenges for this task in a reinforcement learning setting: the mode collapse, the delayed feedback, and the time-consuming warm-up for policy networks. Generally, we propose a novel Curiosity-driven Reinforcement Learning (CRL) framework to jointly enhance the diversity and accuracy of the generated paragraphs. First, by modeling the paragraph captioning as a long-term decision-making process and measuring the prediction uncertainty of state transitions as intrinsic rewards, the model is incentivized to memorize precise but rarely spotted descriptions to context, rather than being biased towards frequent fragments and generic patterns. Second, since the extrinsic reward from evaluation is only available until the complete paragraph is generated, we estimate its expected value at each time step with temporal-difference learning, by considering the correlations between successive actions. Then the estimated extrinsic rewards are complemented by dense intrinsic rewards produced from the derived curiosity module, in order to encourage the policy to fully explore action space and find a global optimum. Third, discounted imitation learning is integrated for learning from human demonstrations, without separately performing the time-consuming warm-up in advance. Extensive experiments conducted on the Standford image-paragraph dataset demonstrate the effectiveness and efficiency of the proposed method, improving the performance by 38.4% compared with state-of-the-art.

Existing deep hashing approaches fail to fully explore semantic correlations and neglect the effect of linguistic context on visual attention learning, leading to inferior performance. This paper proposes a dual-stream learning framework, dubbed Deep Collaborative Discrete Hashing (DCDH), which constructs a discriminative common discrete space by collaboratively incorporating the shared and individual semantics deduced from visual features and semantic labels. Specifically, the context-aware representations are generated by employing the outer product of visual embeddings and semantic encodings. Moreover, we reconstruct the labels and introduce the focal loss to take advantage of frequent and rare concepts. The common binary code space is built on the joint learning of the visual representations attended by language, the semantic-invariant structure construction and the label distribution correction. Extensive experiments demonstrate the superiority of our method.

Rich high-quality annotated data is critical for semantic segmentation learning, yet acquiring dense and pixel-wise ground-truth is both labor- and time-consuming. Coarse annotations (e.g., scribbles, coarse polygons) offer an economical alternative, with which training phase could hardly generate satisfactory performance unfortunately. In order to generate high-quality annotated data with a low time cost for accurate segmentation, in this paper, we propose a novel annotation enrichment strategy, which expands existing coarse annotations of training data to a finer scale. Extensive experiments on the Cityscapes and PASCAL VOC 2012 benchmarks have shown that the neural networks trained with the enriched annotations from our framework yield a significant improvement over that trained with the original coarse labels. It is highly competitive to the performance obtained by using human annotated dense annotations. The proposed method also outperforms among other state-of-the-art weakly-supervised segmentation methods.

With the widespread availability of image captioning at a sentence level, how to automatically generate image paragraphs is yet well explored. Describing an image by a full paragraph involves organising sentences orderly, coherently and diversely, inevitably leading higher complexity than by a single sentence. Existing image paragraph captioning methods give a series of sentences to represent the objects and regions of interests, where the descriptions are essentially generated by feeding the image fragments containing objects and regions into conventional image single-sentence captioning models. This strategy is difficult to generate the descriptions that guarantee the stereoscopic hierarchy and non-overlapping objects. In this paper, we propose a Depth-aware Attention Model (\textitDAM ) to generate paragraph captions for images. The depths of image areas are firstly estimated in order to discriminate objects in a range of spatial locations, which can further guide the linguistic decoder to reveal spatial relationships among objects. This model completes the paragraph in a logical and coherent manner. By incorporating the attention mechanism, the learned model swiftly shifts the sentence focus during paragraph generation, whilst avoiding verbose descriptions on a same object. Extensive quantitative experiments and the user study have been conducted on the Visual Genome dataset, which demonstrate the effectiveness and the interpretability of the proposed model.

To mitigate air pollution problems and reduce greenhouse gas emissions (GHG), plug-in hybrid electric vehicles (PHEV) have been developed to achieve higher fuel efficiency. The Energy Management System (EMS) is a very important component of a PHEV in achieving better fuel economy and it is a very active research area. So far, most of the existing EMS strategies just simple follow predefined rules that are not adaptive to changing driving conditions; other strategies as starting to incorporate accurate prediction of future traffic conditions. In this study, a deep reinforcement learning based PHEV energy management system is designed to autonomously learn the optimal fuel use from its own historical driving record. It is a fully data-driven and learning-enabled model that does not rely on any prediction or predefined rules. The experiment results show that the proposed model is able to achieve 16.3% energy savings comparing to conventional binary control strategies.

Hashing has shown its efficiency and effectiveness in facilitating large-scale multimedia applications. Supervised knowledge (\emph{e.g.}, semantic labels or pair-wise relationship) associated to data is capable of significantly improving the quality of hash codes and hash functions. However, confronted with the rapid growth of newly-emerging concepts and multimedia data on the Web, existing supervised hashing approaches may easily suffer from the scarcity and validity of supervised information due to the expensive cost of manual labelling. In this paper, we propose a novel hashing scheme, termed \emph{zero-shot hashing} (ZSH), which compresses images of "unseen" categories to binary codes with hash functions learned from limited training data of "seen" categories. Specifically, we project independent data labels (i.e., 0/1-form label vectors) into semantic embedding space, where semantic relationships among all the labels can be precisely characterized and thus seen supervised knowledge can be transferred to unseen classes. Moreover, in order to cope with the semantic shift problem, we rotate the embedded space to more suitably align the embedded semantics with the low-level visual feature space, thereby alleviating the influence of semantic gap. In the meantime, to exert positive effects on learning high-quality hash functions, we further propose to preserve local structural property and discrete nature in binary codes. Besides, we develop an efficient alternating algorithm to solve the ZSH model. Extensive experiments conducted on various real-life datasets show the superior zero-shot image retrieval performance of ZSH as compared to several state-of-the-art hashing methods.