Junyu Gao

Tao Han, Lei Bai, Junyu Gao et al.

Pedestrian counting is a fundamental tool for understanding pedestrian patterns and crowd flow analysis. Existing works (e.g., image-level pedestrian counting, crossline crowd counting et al.) either only focus on the image-level counting or are constrained to the manual annotation of lines. In this work, we propose to conduct the pedestrian counting from a new perspective - Video Individual Counting (VIC), which counts the total number of individual pedestrians in the given video (a person is only counted once). Instead of relying on the Multiple Object Tracking (MOT) techniques, we propose to solve the problem by decomposing all pedestrians into the initial pedestrians who existed in the first frame and the new pedestrians with separate identities in each following frame. Then, an end-to-end Decomposition and Reasoning Network (DRNet) is designed to predict the initial pedestrian count with the density estimation method and reason the new pedestrian's count of each frame with the differentiable optimal transport. Extensive experiments are conducted on two datasets with congested pedestrians and diverse scenes, demonstrating the effectiveness of our method over baselines with great superiority in counting the individual pedestrians. Code: https://github.com/taohan10200/DRNet.

Jiwei Chen, Qi Wang, Junyu Gao et al.

Crowd scenes captured by cameras at different locations vary greatly, and existing crowd models have limited generalization for unseen surveillance scenes. To improve the generalization of the model, we regard different surveillance scenes as different category scenes, and introduce few-shot learning to make the model adapt to the unseen surveillance scene that belongs to the given exemplar category scene. To this end, we propose to leverage local and global density characteristics to guide the model of crowd counting for unseen surveillance scenes. Specifically, to enable the model to adapt to the varying density variations in the target scene, we propose the multiple local density learner to learn multi prototypes which represent different density distributions in the support scene. Subsequently, these multiple local density similarity matrixes are encoded. And they are utilized to guide the model in a local way. To further adapt to the global density in the target scene, the global density features are extracted from the support image, then it is used to guide the model in a global way. Experiments on three surveillance datasets shows that proposed method can adapt to the unseen surveillance scene and outperform recent state-of-the-art methods in the few-shot crowd counting.

Junyu Gao, Xuan Yao, Changsheng Xu

The ability to accurately comprehend natural language instructions and navigate to the target location is essential for an embodied agent. Such agents are typically required to execute user instructions in an online manner, leading us to explore the use of unlabeled test samples for effective online model adaptation. However, for online Vision-and-Language Navigation (VLN), due to the intrinsic nature of inter-sample online instruction execution and intra-sample multi-step action decision, frequent updates can result in drastic changes in model parameters, while occasional updates can make the model ill-equipped to handle dynamically changing environments. Therefore, we propose a Fast-Slow Test-Time Adaptation (FSTTA) approach for online VLN by performing joint decomposition-accumulation analysis for both gradients and parameters in a unified framework. Extensive experiments show that our method obtains impressive performance gains on four popular benchmarks. Code is available at https://github.com/Feliciaxyao/ICML2024-FSTTA.

Ziyue Wu, Junyu Gao, Shucheng Huang et al.

Grounding temporal video segments described in natural language queries effectively and efficiently is a crucial capability needed in vision-and-language fields. In this paper, we deal with the fast video temporal grounding (FVTG) task, aiming at localizing the target segment with high speed and favorable accuracy. Most existing approaches adopt elaborately designed cross-modal interaction modules to improve the grounding performance, which suffer from the test-time bottleneck. Although several common space-based methods enjoy the high-speed merit during inference, they can hardly capture the comprehensive and explicit relations between visual and textual modalities. In this paper, to tackle the dilemma of speed-accuracy tradeoff, we propose a commonsense-aware cross-modal alignment (CCA) framework, which incorporates commonsense-guided visual and text representations into a complementary common space for fast video temporal grounding. Specifically, the commonsense concepts are explored and exploited by extracting the structural semantic information from a language corpus. Then, a commonsense-aware interaction module is designed to obtain bridged visual and text features by utilizing the learned commonsense concepts. Finally, to maintain the original semantic information of textual queries, a cross-modal complementary common space is optimized to obtain matching scores for performing FVTG. Extensive results on two challenging benchmarks show that our CCA method performs favorably against state-of-the-arts while running at high speed. Our code is available at https://github.com/ZiyueWu59/CCA.

Jie Fu, Junyu Gao, Changsheng Xu · mila

Weakly-supervised audio-visual video parsing (WS-AVVP) aims to localize the temporal extents of audio, visual and audio-visual event instances as well as identify the corresponding event categories with only video-level category labels for training. Most previous methods pay much attention to refining the supervision for each modality or extracting fruitful cross-modality information for more reliable feature learning. None of them have noticed the imbalanced feature learning between different modalities in the task. In this paper, to balance the feature learning processes of different modalities, a dynamic gradient modulation (DGM) mechanism is explored, where a novel and effective metric function is designed to measure the imbalanced feature learning between audio and visual modalities. Furthermore, principle analysis indicates that the multimodal confusing calculation will hamper the precise measurement of multimodal imbalanced feature learning, which further weakens the effectiveness of our DGM mechanism. To cope with this issue, a modality-separated decision unit (MSDU) is designed for more precise measurement of imbalanced feature learning between audio and visual modalities. Comprehensive experiments are conducted on public benchmarks and the corresponding experimental results demonstrate the effectiveness of our proposed method.

Bingyu Li, Da Zhang, Zhiyuan Zhao et al.

Multimodal semantic segmentation shows significant potential for enhancing segmentation accuracy in complex scenes. However, current methods often incorporate specialized feature fusion modules tailored to specific modalities, thereby restricting input flexibility and increasing the number of training parameters. To address these challenges, we propose StitchFusion, a straightforward yet effective modal fusion framework that integrates large-scale pre-trained models directly as encoders and feature fusers. This approach facilitates comprehensive multi-modal and multi-scale feature fusion, accommodating any visual modal inputs. Specifically, Our framework achieves modal integration during encoding by sharing multi-modal visual information. To enhance information exchange across modalities, we introduce a multi-directional adapter module (MultiAdapter) to enable cross-modal information transfer during encoding. By leveraging MultiAdapter to propagate multi-scale information across pre-trained encoders during the encoding process, StitchFusion achieves multi-modal visual information integration during encoding. Extensive comparative experiments demonstrate that our model achieves state-of-the-art performance on four multi-modal segmentation datasets with minimal additional parameters. Furthermore, the experimental integration of MultiAdapter with existing Feature Fusion Modules (FFMs) highlights their complementary nature. Our code is available at StitchFusion_repo.

Da Zhang, Bingyu Li, Zhiyuan Zhao et al.

Medical time series are central to healthcare, enabling continuous monitoring and supporting timely clinical decisions. Despite recent progress, existing methods struggle to jointly model local-global dynamics and handle nonstationarities like baseline drift, while often failing to capture latent channel interactions. To address these challenges, we propose MedMamba, an end-to-end architecture that integrates state space models with domain-specific inductive biases. Specifically, MedMamba first employs multi-scale convolutional embeddings to capture discriminative local morphology. Second, to mitigate nonstationarity, we introduce a tri-branch differential state space encoder that processes raw, temporal-difference, and frequency-domain views, fusing them to emphasize informative patterns while suppressing drift. Furthermore, to uncover latent channel correlations, we design a spatial graph Mamba module that learns a directed dependency structure regularized toward sparsity and acyclicity, which obviates the need for predefined graphs. Extensive experiments on five real-world datasets demonstrate that MedMamba achieves state-of-the-art performance while maintaining linear computational complexity, and ablation studies validate each component's contribution.Code is available at https://github.com/zhangda1018/MedMamba.

Sikai Bai, Shuaicheng Li, Weiming Zhuang et al.

Federated learning has become a popular method to learn from decentralized heterogeneous data. Federated semi-supervised learning (FSSL) emerges to train models from a small fraction of labeled data due to label scarcity on decentralized clients. Existing FSSL methods assume independent and identically distributed (IID) labeled data across clients and consistent class distribution between labeled and unlabeled data within a client. This work studies a more practical and challenging scenario of FSSL, where data distribution is different not only across clients but also within a client between labeled and unlabeled data. To address this challenge, we propose a novel FSSL framework with dual regulators, FedDure. FedDure lifts the previous assumption with a coarse-grained regulator (C-reg) and a fine-grained regulator (F-reg): C-reg regularizes the updating of the local model by tracking the learning effect on labeled data distribution; F-reg learns an adaptive weighting scheme tailored for unlabeled instances in each client. We further formulate the client model training as bi-level optimization that adaptively optimizes the model in the client with two regulators. Theoretically, we show the convergence guarantee of the dual regulators. Empirically, we demonstrate that FedDure is superior to the existing methods across a wide range of settings, notably by more than 11 on CIFAR-10 and CINIC-10 datasets.

Juncheng Wang, Junyu Gao, Yuan Yuan et al.

Crowd localization is to predict each instance head position in crowd scenarios. Since the distance of instances being to the camera are variant, there exists tremendous gaps among scales of instances within an image, which is called the intrinsic scale shift. The core reason of intrinsic scale shift being one of the most essential issues in crowd localization is that it is ubiquitous in crowd scenes and makes scale distribution chaotic. To this end, the paper concentrates on access to tackle the chaos of the scale distribution incurred by intrinsic scale shift. We propose Gaussian Mixture Scope (GMS) to regularize the chaotic scale distribution. Concretely, the GMS utilizes a Gaussian mixture distribution to adapt to scale distribution and decouples the mixture model into sub-normal distributions to regularize the chaos within the sub-distributions. Then, an alignment is introduced to regularize the chaos among sub-distributions. However, despite that GMS is effective in regularizing the data distribution, it amounts to dislodging the hard samples in training set, which incurs overfitting. We assert that it is blamed on the block of transferring the latent knowledge exploited by GMS from data to model. Therefore, a Scoped Teacher playing a role of bridge in knowledge transform is proposed. What' s more, the consistency regularization is also introduced to implement knowledge transform. To that effect, the further constraints are deployed on Scoped Teacher to derive feature consistence between teacher and student end. With proposed GMS and Scoped Teacher implemented on five mainstream datasets of crowd localization, the extensive experiments demonstrate the superiority of our work. Moreover, comparing with existing crowd locators, our work achieves state-of-the-art via F1-meansure comprehensively on five datasets.

Bingyu Li, Da Zhang, Tao Huo et al.

Large Vision-Language Models (LVLMs) have shown strong visual understanding and language-guided grounding abilities, yet their capacity for multi-temporal visual reasoning remains underexplored. To bridge this gap, we introduce \textbf{Multi-temporal Referring Segmentation (MTRS)}, a new task that aims to segment language-described temporal changes from multi-temporal images. MTRS extends conventional referring segmentation and change detection by jointly requiring temporal correspondence reasoning, language grounding, and pixel-level mask prediction. We propose \textbf{CRAFT-Agent}, an automated data construction pipeline with human auditing, and build \textbf{MTRefSeg-21K}, the first MTRS benchmark, containing 21K high-quality multi-temporal image-text-mask triplets across diverse scenes, viewpoints, and domains. Benchmarking a broad set of VLM- and LVLM-based models reveals that direct inference performs poorly, while task-specific fine-tuning remains limited. To address this, we propose \textbf{MTRefSeg-R1}, a change-aware LVLM framework trained with a two-stage strategy. It first learns general temporal-change perception from 20K vision-only bi-temporal samples, and is then fine-tuned on MTRefSeg-21K for fine-grained language-guided temporal localization. MTRefSeg-R1 explicitly models cross-temporal visual differences, aligns language instructions with temporal variations, and predicts referred change masks. Extensive experiments show that MTRefSeg-R1 achieves strong and often superior performance compared with existing LVLM baselines, demonstrating the challenge and potential of MTRS.

Bingyu Li, Tao Huo, Haocheng Dong et al.

Open-vocabulary remote sensing image segmentation (OVRSIS) remains underexplored due to fragmented datasets, limited training diversity, and the lack of evaluation benchmarks that reflect realistic geospatial application demands. Our previous \textit{OVRSISBenchV1} established an initial cross-dataset evaluation protocol, but its limited scope is insufficient for assessing realistic open-world generalization. To address this issue, we propose \textit{OVRSISBenchV2}, a large-scale and application-oriented benchmark for OVRSIS. We first construct \textbf{OVRSIS95K}, a balanced dataset of about 95K image--mask pairs covering 35 common semantic categories across diverse remote sensing scenes. Built upon OVRSIS95K and 10 downstream datasets, OVRSISBenchV2 contains 170K images and 128 categories, substantially expanding scene diversity, semantic coverage, and evaluation difficulty. Beyond standard open-vocabulary segmentation, it further includes downstream protocols for building extraction, road extraction, and flood detection, thereby better reflecting realistic geospatial application demands and complex deployment scenarios. We also propose \textbf{Pi-Seg}, a baseline for OVRSIS. Pi-Seg improves transferability through a \textbf{positive-incentive noise} mechanism, where learnable and semantically guided perturbations broaden the visual-text feature space during training. Extensive experiments on OVRSISBenchV1, OVRSISBenchV2, and downstream tasks show that Pi-Seg delivers strong and consistent results, particularly on the more challenging OVRSISBenchV2 benchmark. Our results highlight both the importance of realistic benchmark design and the effectiveness of perturbation-based transfer for OVRSIS. The code and datasets are available at \href{https://github.com/LiBingyu01/RSKT-Seg/tree/Pi-Seg}{LiBingyu01/RSKT-Seg/tree/Pi-Seg}.

Yaoteng Zhang, Qing Zhou, Junyu Gao et al.

Remote sensing imagery typically arrives in the form of continuous data streams. Traditional detectors often forget previously learned categories when learning new ones; therefore, research on Remote Sensing Incremental Object Detection (RS-IOD) is of great significance. However, existing methods largely overlook the intra-class scale variations prevalent in remote sensing scenes, which undermines the effectiveness of knowledge transfer and old knowledge preservation. Moreover, RS-IOD also suffers from missing annotations, which cause the model to misclassify old-class instances as background. To address these challenges, we propose a novel framework, STAR-IOD. First, we introduce a Subspace-decoupled Topology Distillation (STD) module to transfer structural knowledge, explicitly aligning inter-class topological relationships and mitigating intra-class representation discrepancies induced by scale shifts. Furthermore, we introduce the Clustering-driven Pseudo-label Generator (CPG), a plug-and-play module that leverages K-Means clustering to dynamically identify class-specific thresholds, thereby guaranteeing an accurate distinction between true positive targets and background noise and alleviating the issue of missing annotations for old classes. We also constructed two Remote Sensing Incremental Object Detection datasets, DIOR-IOD and DOTA-IOD to facilitate research on RS-IOD. Extensive experiments demonstrate that our method outperforms state-of-the-art approaches by 1.7% and 2.1% mAP on DIOR-IOD and DOTA-IOD, respectively, effectively alleviating catastrophic forgetting while preserving strong detection performance on both base and novel classes. The code and dataset are released at: https://github.com/zyt95579/STAR-IOD.

Shengkai Sun, Daizong Liu, Jianfeng Dong et al.

Unsupervised pre-training has shown great success in skeleton-based action understanding recently. Existing works typically train separate modality-specific models, then integrate the multi-modal information for action understanding by a late-fusion strategy. Although these approaches have achieved significant performance, they suffer from the complex yet redundant multi-stream model designs, each of which is also limited to the fixed input skeleton modality. To alleviate these issues, in this paper, we propose a Unified Multimodal Unsupervised Representation Learning framework, called UmURL, which exploits an efficient early-fusion strategy to jointly encode the multi-modal features in a single-stream manner. Specifically, instead of designing separate modality-specific optimization processes for uni-modal unsupervised learning, we feed different modality inputs into the same stream with an early-fusion strategy to learn their multi-modal features for reducing model complexity. To ensure that the fused multi-modal features do not exhibit modality bias, i.e., being dominated by a certain modality input, we further propose both intra- and inter-modal consistency learning to guarantee that the multi-modal features contain the complete semantics of each modal via feature decomposition and distinct alignment. In this manner, our framework is able to learn the unified representations of uni-modal or multi-modal skeleton input, which is flexible to different kinds of modality input for robust action understanding in practical cases. Extensive experiments conducted on three large-scale datasets, i.e., NTU-60, NTU-120, and PKU-MMD II, demonstrate that UmURL is highly efficient, possessing the approximate complexity with the uni-modal methods, while achieving new state-of-the-art performance across various downstream task scenarios in skeleton-based action representation learning.

Bingyu Li, Da Zhang, Zhiyuan Zhao et al.

Text has become the predominant form of communication on social media, embedding a wealth of emotional nuances. Consequently, the extraction of emotional information from text is of paramount importance. Despite previous research making some progress, existing text sentiment analysis models still face challenges in integrating diverse semantic information and lack interpretability. To address these issues, we propose a quantum-inspired deep learning architecture that combines fundamental principles of quantum mechanics (QM principles) with deep learning models for text sentiment analysis. Specifically, we analyze the commonalities between text representation and QM principles to design a quantum-inspired text representation method and further develop a quantum-inspired text embedding layer. Additionally, we design a feature extraction layer based on long short-term memory (LSTM) networks and self-attention mechanisms (SAMs). Finally, we calculate the text density matrix using the quantum complex numbers principle and apply 2D-convolution neural networks (CNNs) for feature condensation and dimensionality reduction. Through a series of visualization, comparative, and ablation experiments, we demonstrate that our model not only shows significant advantages in accuracy and efficiency compared to previous related models but also achieves a certain level of interpretability by integrating QM principles. Our code is available at QISA.

Qi Wang, Juncheng Wang, Junyu Gao et al.

The mainstream crowd counting methods regress density map and integrate it to obtain counting results. Since the density representation to one head accords to its adjacent distribution, it embeds the same category objects with variant values, while human beings counting models the invariant features namely similarity to objects. Inspired by this, we propose a rational and anthropoid crowd counting framework. To begin with, we leverage counting scalar as supervision signal, which provides global and implicit guidance to similar matters. Then, the large kernel CNN is utilized to imitate the paradigm of human beings which models invariant knowledge firstly and slides to compare similarity. Later, re-parameterization on pre-trained paralleled parameters is presented to cater to the inner-class variance on similarity comparison. Finally, the Random Scaling patches Yield (RSY) is proposed to facilitate similarity modeling on long distance dependencies. Extensive experiments on five challenging benchmarks in crowd counting show the proposed framework achieves state-of-the-art.

Da Zhang, Bingyu Li, Zhiyuan Zhao et al.

Time series analysis plays a vital role in fields such as finance, healthcare, industry, and meteorology, underpinning key tasks including classification, forecasting, and anomaly detection. Although deep learning models have achieved remarkable progress in these areas in recent years, constructing an efficient, multi-task compatible, and generalizable unified framework for time series analysis remains a significant challenge. Existing approaches are often tailored to single tasks or specific data types, making it difficult to simultaneously handle multi-task modeling and effectively integrate information across diverse time series types. Moreover, real-world data are often affected by noise, complex frequency components, and multi-scale dynamic patterns, which further complicate robust feature extraction and analysis. To ameliorate these challenges, we propose FusAD, a unified analysis framework designed for diverse time series tasks. FusAD features an adaptive time-frequency fusion mechanism, integrating both Fourier and Wavelet transforms to efficiently capture global-local and multi-scale dynamic features. With an adaptive denoising mechanism, FusAD automatically senses and filters various types of noise, highlighting crucial sequence variations and enabling robust feature extraction in complex environments. In addition, the framework integrates a general information fusion and decoding structure, combined with masked pre-training, to promote efficient learning and transfer of multi-granularity representations. Extensive experiments demonstrate that FusAD consistently outperforms state-of-the-art models on mainstream time series benchmarks for classification, forecasting, and anomaly detection tasks, while maintaining high efficiency and scalability. Code is available at https://github.com/zhangda1018/FusAD.

Haoxuan Ding, Qi Wang, Junyu Gao et al.

Traditional license plate detection and recognition models are often trained on closed datasets, limiting their ability to handle the diverse license plate formats across different regions. The emergence of large-scale pre-trained models has shown exceptional generalization capabilities, enabling few-shot and zero-shot learning. We propose OneShotLP, a training-free framework for video-based license plate detection and recognition, leveraging these advanced models. Starting with the license plate position in the first video frame, our method tracks this position across subsequent frames using a point tracking module, creating a trajectory of prompts. These prompts are input into a segmentation module that uses a promptable large segmentation model to generate local masks of the license plate regions. The segmented areas are then processed by multimodal large language models (MLLMs) for accurate license plate recognition. OneShotLP offers significant advantages, including the ability to function effectively without extensive training data and adaptability to various license plate styles. Experimental results on UFPR-ALPR and SSIG-SegPlate datasets demonstrate the superior accuracy of our approach compared to traditional methods. This highlights the potential of leveraging pre-trained models for diverse real-world applications in intelligent transportation systems. The code is available at https://github.com/Dinghaoxuan/OneShotLP.

Junyu Gao, Mengyuan Chen, Liangyu Xiang et al.

Reliable uncertainty estimation has become a crucial requirement for the industrial deployment of deep learning algorithms, particularly in high-risk applications such as autonomous driving and medical diagnosis. However, mainstream uncertainty estimation methods, based on deep ensembling or Bayesian neural networks, generally impose substantial computational overhead. To address this challenge, a novel paradigm called Evidential Deep Learning (EDL) has emerged, providing reliable uncertainty estimation with minimal additional computation in a single forward pass. This survey provides a comprehensive overview of the current research on EDL, designed to offer readers a broad introduction to the field without assuming prior knowledge. Specifically, we first delve into the theoretical foundation of EDL, the subjective logic theory, and discuss its distinctions from other uncertainty estimation frameworks. We further present existing theoretical advancements in EDL from four perspectives: reformulating the evidence collection process, improving uncertainty estimation via OOD samples, delving into various training strategies, and evidential regression networks. Thereafter, we elaborate on its extensive applications across various machine learning paradigms and downstream tasks. In the end, an outlook on future directions for better performances and broader adoption of EDL is provided, highlighting potential research avenues.

Junyu Gao, Xinhong Ma, Changsheng Xu

Despite the great progress of unsupervised domain adaptation (UDA) with the deep neural networks, current UDA models are opaque and cannot provide promising explanations, limiting their applications in the scenarios that require safe and controllable model decisions. At present, a surge of work focuses on designing deep interpretable methods with adequate data annotations and only a few methods consider the distributional shift problem. Most existing interpretable UDA methods are post-hoc ones, which cannot facilitate the model learning process for performance enhancement. In this paper, we propose an inherently interpretable method, named Transferable Conceptual Prototype Learning (TCPL), which could simultaneously interpret and improve the processes of knowledge transfer and decision-making in UDA. To achieve this goal, we design a hierarchically prototypical module that transfers categorical basic concepts from the source domain to the target domain and learns domain-shared prototypes for explaining the underlying reasoning process. With the learned transferable prototypes, a self-predictive consistent pseudo-label strategy that fuses confidence, predictions, and prototype information, is designed for selecting suitable target samples for pseudo annotations and gradually narrowing down the domain gap. Comprehensive experiments show that the proposed method can not only provide effective and intuitive explanations but also outperform previous state-of-the-arts.

Yufan Hu, Junyu Gao, Changsheng Xu

Most existing state-of-the-art video classification methods assume that the training data obey a uniform distribution. However, video data in the real world typically exhibit an imbalanced long-tailed class distribution, resulting in a model bias towards head class and relatively low performance on tail class. While the current long-tailed classification methods usually focus on image classification, adapting it to video data is not a trivial extension. We propose an end-to-end multi-expert distribution calibration method to address these challenges based on two-level distribution information. The method jointly considers the distribution of samples in each class (intra-class distribution) and the overall distribution of diverse data (inter-class distribution) to solve the issue of imbalanced data under long-tailed distribution. By modeling the two-level distribution information, the model can jointly consider the head classes and the tail classes and significantly transfer the knowledge from the head classes to improve the performance of the tail classes. Extensive experiments verify that our method achieves state-of-the-art performance on the long-tailed video classification task.

Pengyu Chen, Junyu Gao, Yuan Yuan et al.

RGB-Thermal (RGB-T) crowd counting is a challenging task, which uses thermal images as complementary information to RGB images to deal with the decreased performance of unimodal RGB-based methods in scenes with low-illumination or similar backgrounds. Most existing methods propose well-designed structures for cross-modal fusion in RGB-T crowd counting. However, these methods have difficulty in encoding cross-modal contextual semantic information in RGB-T image pairs. Considering the aforementioned problem, we propose a two-stream RGB-T crowd counting network called Multi-Attention Fusion Network (MAFNet), which aims to fully capture long-range contextual information from the RGB and thermal modalities based on the attention mechanism. Specifically, in the encoder part, a Multi-Attention Fusion (MAF) module is embedded into different stages of the two modality-specific branches for cross-modal fusion at the global level. In addition, a Multi-modal Multi-scale Aggregation (MMA) regression head is introduced to make full use of the multi-scale and contextual information across modalities to generate high-quality crowd density maps. Extensive experiments on two popular datasets show that the proposed MAFNet is effective for RGB-T crowd counting and achieves the state-of-the-art performance.

Yaoteng Zhang, Zhou Qing, Junyu Gao et al.

Incremental Object Detection (IOD) aims to continuously learn new object categories without forgetting previously learned ones. Recently, prompt-based methods have gained popularity for their replay-free design and parameter efficiency. However, due to prompt coupling and prompt drift, these methods often suffer from prompt degradation during continual adaptation. To address these issues, we propose a novel prompt-decoupled framework called PDP. PDP innovatively designs a dual-pool prompt decoupling paradigm, which consists of a shared pool used to capture task-general knowledge for forward transfer, and a private pool used to learn task-specific discriminative features. This paradigm explicitly separates task-general and task-specific prompts, preventing interference between prompts and mitigating prompt coupling. In addition, to counteract prompt drift resulting from inconsistent supervision where old foreground objects are treated as background in subsequent tasks, PDP introduces a Prototypical Pseudo-Label Generation (PPG) module. PPG can dynamically update the class prototype space during training and use the class prototypes to further filter valuable pseudo-labels, maintaining supervisory signal consistency throughout the incremental process. PDP achieves state-of-the-art performance on MS-COCO (with a 9.2\% AP improvement) and PASCAL VOC (with a 3.3\% AP improvement) benchmarks, highlighting its potential in balancing stability and plasticity. The code and dataset are released at: https://github.com/zyt95579/PDP\_IOD/tree/main

Yangguang Lin, Quan Fang, Yufei Li et al.

Object hallucination in Large Vision-Language Models (LVLMs) significantly hinders their reliable deployment. Existing methods struggle to balance efficiency and accuracy: they often require expensive reference models and multiple forward passes, or apply static edits that risk suppressing genuine visual evidence. To address this, we introduce HulluEdit, a single-pass, reference-free intervention framework. Our core innovation is orthogonal subspace editing: we decompose the hidden states of the model into orthogonal subspaces - visual evidence, conflicting priors, and residual uncertainty - enabling selective suppression of hallucinatory patterns without interfering with visual grounding. This approach mathematically guarantees that edits applied to the prior subspace leave the visual component entirely unaffected. Extensive experiments show that HulluEdit achieves state-of-the-art hallucination reduction on benchmarks including POPE and CHAIR across diverse architectures, while preserving general capabilities on MME and maintaining efficient inference. Our method consistently outperforms contrastive decoding and static subspace editing baselines, offering a new pathway toward more trustworthy LVLMs.

Wen Guo, Pengfei Zhao, Zongmeng Wang et al.

Multiple Object Tracking (MOT) has long been a fundamental task in computer vision, with broad applications in various real-world scenarios. However, due to distribution shifts in appearance, motion pattern, and catagory between the training and testing data, model performance degrades considerably during online inference in MOT. Test-Time Adaptation (TTA) has emerged as a promising paradigm to alleviate such distribution shifts. However, existing TTA methods often fail to deliver satisfactory results in MOT, as they primarily focus solely on frame-level adaptation while neglecting temporal consistency and identity association across frames and videos. Inspired by human decision-making process, this paper propose a Test-time Calibration from Experience and Intuition (TCEI) framework. In this framework, the Intuitive system utilizes transient memory to recall recently observed objects for rapid predictions, while the Experiential system leverages the accumulated experience from prior test videos to reassess and calibrate these intuitive predictions. Furthermore, both confident and uncertain objects during online testing are exploited as historical priors and reflective cases, respectively, enabling the model to adapt to the testing environment and alleviate performance degradation. Extensive experiments demonstrate that the proposed TCEI framework consistently achieves superior performance across multiple benchmark datasets and significantly enhances the model's adaptability under distribution shifts. The code will be released at https://github.com/1941Zpf/TCEI.

Xu Han, Junyu Gao, Chuang Yang et al.

The irregular contour representation is one of the tough challenges in scene text detection. Although segmentation-based methods have achieved significant progress with the help of flexible pixel prediction, the overlap of geographically close texts hinders detecting them separately. To alleviate this problem, some shrink-based methods predict text kernels and expand them to restructure texts. However, the text kernel is an artificial object with incomplete semantic features that are prone to incorrect or missing detection. In addition, different from the general objects, the geometry features (aspect ratio, scale, and shape) of scene texts vary significantly, which makes it difficult to detect them accurately. To consider the above problems, we propose an effective spotlight text detector (STD), which consists of a spotlight calibration module (SCM) and a multivariate information extraction module (MIEM). The former concentrates efforts on the candidate kernel, like a camera focus on the target. It obtains candidate features through a mapping filter and calibrates them precisely to eliminate some false positive samples. The latter designs different shape schemes to explore multiple geometric features for scene texts. It helps extract various spatial relationships to improve the model's ability to recognize kernel regions. Ablation studies prove the effectiveness of the designed SCM and MIEM. Extensive experiments verify that our STD is superior to existing state-of-the-art methods on various datasets, including ICDAR2015, CTW1500, MSRA-TD500, and Total-Text.

Bingyu Li, Da Zhang, Zhiyuan Zhao et al.

Multimodal semantic segmentation is a pivotal component of computer vision and typically surpasses unimodal methods by utilizing rich information set from various sources.Current models frequently adopt modality-specific frameworks that inherently biases toward certain modalities. Although these biases might be advantageous in specific situations, they generally limit the adaptability of the models across different multimodal contexts, thereby potentially impairing performance. To address this issue, we leverage the inherent capabilities of the model itself to discover the optimal equilibrium in multimodal fusion and introduce U3M: An Unbiased Multiscale Modal Fusion Model for Multimodal Semantic Segmentation. Specifically, this method involves an unbiased integration of multimodal visual data. Additionally, we employ feature fusion at multiple scales to ensure the effective extraction and integration of both global and local features. Experimental results demonstrate that our approach achieves superior performance across multiple datasets, verifing its efficacy in enhancing the robustness and versatility of semantic segmentation in diverse settings. Our code is available at U3M-multimodal-semantic-segmentation.

Mengyuan Chen, Junyu Gao, Changsheng Xu

A straightforward pipeline for zero-shot out-of-distribution (OOD) detection involves selecting potential OOD labels from an extensive semantic pool and then leveraging a pre-trained vision-language model to perform classification on both in-distribution (ID) and OOD labels. In this paper, we theorize that enhancing performance requires expanding the semantic pool, while increasing the expected probability of selected OOD labels being activated by OOD samples, and ensuring low mutual dependence among the activations of these OOD labels. A natural expansion manner is to adopt a larger lexicon; however, the inevitable introduction of numerous synonyms and uncommon words fails to meet the above requirements, indicating that viable expansion manners move beyond merely selecting words from a lexicon. Since OOD detection aims to correctly classify input images into ID/OOD class groups, we can "make up" OOD label candidates which are not standard class names but beneficial for the process. Observing that the original semantic pool is comprised of unmodified specific class names, we correspondingly construct a conjugated semantic pool (CSP) consisting of modified superclass names, each serving as a cluster center for samples sharing similar properties across different categories. Consistent with our established theory, expanding OOD label candidates with the CSP satisfies the requirements and outperforms existing works by 7.89% in FPR95. Codes are available in https://github.com/MengyuanChen21/NeurIPS2024-CSP.

Haoxuan Ding, Junyu Gao, Yuan Yuan et al.

With the emergence of foundation model, this novel paradigm of deep learning has encouraged many powerful achievements in natural language processing and computer vision. There are many advantages of foundation model, such as excellent feature extraction power, mighty generalization ability, great few-shot and zero-shot learning capacity, etc. which are beneficial to vision tasks. As the unique identity of vehicle, different countries and regions have diverse license plate (LP) styles and appearances, and even different types of vehicles have different LPs. However, recent deep learning based license plate detectors are mainly trained on specific datasets, and these limited datasets constrain the effectiveness and robustness of LP detectors. To alleviate the negative impact of limited data, an attempt to exploit the advantages of foundation model is implement in this paper. We customize a vision foundation model, i.e. Segment Anything Model (SAM), for LP detection task and propose the first LP detector based on vision foundation model, named SamLP. Specifically, we design a Low-Rank Adaptation (LoRA) fine-tuning strategy to inject extra parameters into SAM and transfer SAM into LP detection task. And then, we further propose a promptable fine-tuning step to provide SamLP with prompatable segmentation capacity. The experiments show that our proposed SamLP achieves promising detection performance compared to other LP detectors. Meanwhile, the proposed SamLP has great few-shot and zero-shot learning ability, which shows the potential of transferring vision foundation model. The code is available at https://github.com/Dinghaoxuan/SamLP

Xuan Yao, Junyu Gao, Changsheng Xu

Vision-and-Language Navigation in Continuous Environments (VLN-CE) requires agents to execute sequential navigation actions in complex environments guided by natural language instructions. Current approaches often struggle with generalizing to novel environments and adapting to ongoing changes during navigation. Inspired by human cognition, we present NavMorph, a self-evolving world model framework that enhances environmental understanding and decision-making in VLN-CE tasks. NavMorph employs compact latent representations to model environmental dynamics, equipping agents with foresight for adaptive planning and policy refinement. By integrating a novel Contextual Evolution Memory, NavMorph leverages scene-contextual information to support effective navigation while maintaining online adaptability. Extensive experiments demonstrate that our method achieves notable performance improvements on popular VLN-CE benchmarks. Code is available at https://github.com/Feliciaxyao/NavMorph.

Xu Han, Junyu Gao, Chuang Yang et al.

Due to the diversity of scene text in aspects such as font, color, shape, and size, accurately and efficiently detecting text is still a formidable challenge. Among the various detection approaches, segmentation-based approaches have emerged as prominent contenders owing to their flexible pixel-level predictions. However, these methods typically model text instances in a bottom-up manner, which is highly susceptible to noise. In addition, the prediction of pixels is isolated without introducing pixel-feature interaction, which also influences the detection performance. To alleviate these problems, we propose a multi-information level arbitrary-shaped text detector consisting of a focus entirety module (FEM) and a perceive environment module (PEM). The former extracts instance-level features and adopts a top-down scheme to model texts to reduce the influence of noises. Specifically, it assigns consistent entirety information to pixels within the same instance to improve their cohesion. In addition, it emphasizes the scale information, enabling the model to distinguish varying scale texts effectively. The latter extracts region-level information and encourages the model to focus on the distribution of positive samples in the vicinity of a pixel, which perceives environment information. It treats the kernel pixels as positive samples and helps the model differentiate text and kernel features. Extensive experiments demonstrate the FEM's ability to efficiently support the model in handling different scale texts and confirm the PEM can assist in perceiving pixels more accurately by focusing on pixel vicinities. Comparisons show the proposed model outperforms existing state-of-the-art approaches on four public datasets.

Da Zhang, Bingyu Li, Feiyu Wang et al.

Zero-shot object counting (ZSOC) aims to enumerate objects of arbitrary categories specified by text descriptions without requiring visual exemplars. However, existing methods often treat counting as a coarse retrieval task, suffering from a lack of fine-grained quantity awareness. Furthermore, they frequently exhibit spatial insensitivity and degraded generalization due to feature space distortion during model adaptation.To address these challenges, we present \textbf{QICA}, a novel framework that synergizes \underline{q}uantity percept\underline{i}on with robust spatial \underline{c}ast \underline{a}ggregation. Specifically, we introduce a Synergistic Prompting Strategy (\textbf{SPS}) that adapts vision and language encoders through numerically conditioned prompts, bridging the gap between semantic recognition and quantitative reasoning. To mitigate feature distortion, we propose a Cost Aggregation Decoder (\textbf{CAD}) that operates directly on vision-text similarity maps. By refining these maps through spatial aggregation, CAD prevents overfitting while preserving zero-shot transferability. Additionally, a multi-level quantity alignment loss ($\mathcal{L}_{MQA}$) is employed to enforce numerical consistency across the entire pipeline. Extensive experiments on FSC-147 demonstrate competitive performance, while zero-shot evaluation on CARPK and ShanghaiTech-A validates superior generalization to unseen domains.

Da Zhang, Bingyu Li, Zhuyuan Zhao et al.

As multimodal data proliferates across diverse real-world applications, leveraging heterogeneous information such as texts and timestamps for accurate time series forecasting (TSF) has become a critical challenge. While diffusion models demonstrate exceptional performance in generation tasks, their application to TSF remains largely confined to modeling single-modality numerical sequences, overlooking the abundant cross-modal signals inherent in complex heterogeneous data. To address this gap, we propose UniDiff, a unified diffusion framework for multimodal time series forecasting. To process the numerical sequence, our framework first tokenizes the time series into patches, preserving local temporal dynamics by mapping each patch to an embedding space via a lightweight MLP. At its core lies a unified and parallel fusion module, where a single cross-attention mechanism adaptively weighs and integrates structural information from timestamps and semantic context from texts in one step, enabling a flexible and efficient interplay between modalities. Furthermore, we introduce a novel classifier-free guidance mechanism designed for multi-source conditioning, allowing for decoupled control over the guidance strength of textual and temporal information during inference, which significantly enhances model robustness. Extensive experiments on real-world benchmark datasets across eight domains demonstrate that the proposed UniDiff model achieves state-of-the-art performance.

Siqu Ou, Tianrui Wan, Zhiyuan Zhao et al.

While chain-of-thought (CoT) reasoning has substantially improved multimodal large language models (MLLMs) on complex reasoning tasks, existing approaches largely rely on long textual reasoning trajectories and provide limited mechanisms for learning stable visual attention policies. Our analysis shows that current MLLMs exhibit weak visual focus: early-stage visual misalignment is rarely corrected during subsequent reasoning, leading to error propagation and failed inferences. We argue that this limitation stems from inadequate credit assignment for visual attention during training. To address this issue, we propose SAYO, a visual reasoning model trained with a reinforcement learning (RL) framework that introduces a region-level visual attention-based reward. This reward explicitly aligns optimization signals with visually grounded reasoning steps, enabling the model to learn more reliable attention behaviors. Extensive experiments across multiple multimodal benchmarks demonstrate that SAYO consistently improves performance on diverse reasoning and perception tasks.

Bingyu Li, Tao Huo, Da Zhang et al.

Accurate segmentation of marine organisms is vital for biodiversity monitoring and ecological assessment, yet existing datasets and models remain largely limited to terrestrial scenes. To bridge this gap, we introduce \textbf{AquaOV255}, the first large-scale and fine-grained underwater segmentation dataset containing 255 categories and over 20K images, covering diverse categories for open-vocabulary (OV) evaluation. Furthermore, we establish the first underwater OV segmentation benchmark, \textbf{UOVSBench}, by integrating AquaOV255 with five additional underwater datasets to enable comprehensive evaluation. Alongside, we present \textbf{Earth2Ocean}, a training-free OV segmentation framework that transfers terrestrial vision--language models (VLMs) to underwater domains without any additional underwater training. Earth2Ocean consists of two core components: a Geometric-guided Visual Mask Generator (\textbf{GMG}) that refines visual features via self-similarity geometric priors for local structure perception, and a Category-visual Semantic Alignment (\textbf{CSA}) module that enhances text embeddings through multimodal large language model reasoning and scene-aware template construction. Extensive experiments on the UOVSBench benchmark demonstrate that Earth2Ocean achieves significant performance improvement on average while maintaining efficient inference.

Feiyu Wang, Jiayuan Yang, Zhiyuan Zhao et al.

Scalable Vector Graphics (SVG) are central to digital design due to their inherent scalability and editability. Despite significant advancements in content generation enabled by Visual Language Models (VLMs), existing text-to-SVG generation methods are limited by a core challenge: the autoregressive training process does not incorporate visual perception of the final rendered image, which fundamentally constrains generation quality. To address this limitation, we propose an Introspective SVG Generation Framework (IntroSVG). At its core, the framework instantiates a unified VLM that operates in a closed loop, assuming dual roles of both generator and critic. Specifically, through Supervised Fine-Tuning (SFT), the model learns to draft SVGs and to provide feedback on their rendered outputs; moreover, we systematically convert early-stage failures into high-quality error-correction training data, thereby enhancing model robustness. Subsequently, we leverage a high-capacity teacher VLM to construct a preference dataset and further align the generator's policy through Direct Preference Optimization (DPO). During inference, the optimized generator and critic operate collaboratively in an iterative "generate-review-refine" cycle, starting from imperfect intermediate drafts to autonomously improve output quality. Experimental results demonstrate that our method achieves state-of-the-art performance across several key evaluation metrics, generating SVGs with more complex structures, stronger semantic alignment, and greater editability. These results corroborate the effectiveness of incorporating explicit visual feedback into the generation loop.

Xu Han, Junyu Gao, Chuang Yang et al.

Texts on the intelligent transportation scene include mass information. Fully harnessing this information is one of the critical drivers for advancing intelligent transportation. Unlike the general scene, detecting text in transportation has extra demand, such as a fast inference speed, except for high accuracy. Most existing real-time text detection methods are based on the shrink mask, which loses some geometry semantic information and needs complex post-processing. In addition, the previous method usually focuses on correct output, which ignores feature correction and lacks guidance during the intermediate process. To this end, we propose an efficient multi-scene text detector that contains an effective text representation similar mask (SM) and a feature correction module (FCM). Unlike previous methods, the former aims to preserve the geometric information of the instances as much as possible. Its post-progressing saves 50$\%$ of the time, accurately and efficiently reconstructing text contours. The latter encourages false positive features to move away from the positive feature center, optimizing the predictions from the feature level. Some ablation studies demonstrate the efficiency of the SM and the effectiveness of the FCM. Moreover, the deficiency of existing traffic datasets (such as the low-quality annotation or closed source data unavailability) motivated us to collect and annotate a traffic text dataset, which introduces motion blur. In addition, to validate the scene robustness of the SM-Net, we conduct experiments on traffic, industrial, and natural scene datasets. Extensive experiments verify it achieves (SOTA) performance on several benchmarks. The code and dataset are available at: \url{https://github.com/fengmulin/SMNet}.

Qing Zhou, Junyu Gao, Qi Wang

The rapid growth of dataset scales has been a key driver in advancing deep learning research. However, as dataset scale increases, the training process becomes increasingly inefficient due to the presence of low-value samples, including excessive redundant samples, overly challenging samples, and inefficient easy samples that contribute little to model improvement.To address this challenge, we propose Scale Efficient Training (SeTa) for large datasets, a dynamic sample pruning approach that losslessly reduces training time. To remove low-value samples, SeTa first performs random pruning to eliminate redundant samples, then clusters the remaining samples according to their learning difficulty measured by loss. Building upon this clustering, a sliding window strategy is employed to progressively remove both overly challenging and inefficient easy clusters following an easy-to-hard curriculum.We conduct extensive experiments on large-scale synthetic datasets, including ToCa, SS1M, and ST+MJ, each containing over 3 million samples.SeTa reduces training costs by up to 50\% while maintaining or improving performance, with minimal degradation even at 70\% cost reduction. Furthermore, experiments on various scale real datasets across various backbones (CNNs, Transformers, and Mambas) and diverse tasks (instruction tuning, multi-view stereo, geo-localization, composed image retrieval, referring image segmentation) demonstrate the powerful effectiveness and universality of our approach. Code is available at https://github.com/mrazhou/SeTa.

Qing Zhou, Tao Yang, Junyu Gao et al.

Remote Sensing Image Captioning (RSIC) is a cross-modal field bridging vision and language, aimed at automatically generating natural language descriptions of features and scenes in remote sensing imagery. Despite significant advances in developing sophisticated methods and large-scale datasets for training vision-language models (VLMs), two critical challenges persist: the scarcity of non-English descriptive datasets and the lack of multilingual capability evaluation for models. These limitations fundamentally impede the progress and practical deployment of RSIC, particularly in the era of large VLMs. To address these challenges, this paper presents several significant contributions to the field. First, we introduce and analyze BRSIC (Bilingual Remote Sensing Image Captioning), a comprehensive bilingual dataset that enriches three established English RSIC datasets with Chinese descriptions, encompassing 13,634 images paired with 68,170 bilingual captions. Building upon this foundation, we develop a systematic evaluation framework that addresses the prevalent inconsistency in evaluation protocols, enabling rigorous assessment of model performance through standardized retraining procedures on BRSIC. Furthermore, we present an extensive empirical study of eight state-of-the-art large vision-language models (LVLMs), examining their capabilities across multiple paradigms including zero-shot inference, supervised fine-tuning, and multi-lingual training. This comprehensive evaluation provides crucial insights into the strengths and limitations of current LVLMs in handling multilingual remote sensing tasks. Additionally, our cross-dataset transfer experiments reveal interesting findings. The code and data will be available at https://github.com/mrazhou/BRSIC.

Tao Hana, Zhibin Wen, Zhenghao Chen et al.

While AI-based numerical weather prediction (NWP) enables rapid forecasting, generating high-resolution outputs remains computationally demanding due to limited multi-scale adaptability and inefficient data representations. We propose the 3D Gaussian splatting-based scale-aware vision transformer (GSSA-ViT), a novel framework for arbitrary-resolution forecasting and flexible downscaling of high-dimensional atmospheric fields. Specifically, latitude-longitude grid points are treated as centers of 3D Gaussians. A generative 3D Gaussian prediction scheme is introduced to estimate key parameters, including covariance, attributes, and opacity, for unseen samples, improving generalization and mitigating overfitting. In addition, a scale-aware attention module is designed to capture cross-scale dependencies, enabling the model to effectively integrate information across varying downscaling ratios and support continuous resolution adaptation. To our knowledge, this is the first NWP approach that combines generative 3D Gaussian modeling with scale-aware attention for unified multi-scale prediction. Experiments on ERA5 show that the proposed method accurately forecasts 87 atmospheric variables at arbitrary resolutions, while evaluations on ERA5 and CMIP6 demonstrate its superior performance in downscaling tasks. The proposed framework provides an efficient and scalable solution for high-resolution, multi-scale atmospheric prediction and downscaling. Code is available at: https://github.com/binbin2xs/weather-GS.

Qing Zhou, Bingxuan Zhao, Tao Yang et al.

Dynamic data pruning accelerates deep learning by selectively omitting less informative samples during training. While per-sample loss is a common importance metric, obtaining it can be challenging or infeasible for complex models or loss functions, often requiring significant implementation effort. This work proposes the Batch Loss Score (BLS), a computationally efficient alternative using an Exponential Moving Average (EMA) of readily available batch losses to assign scores to individual samples. We frame the batch loss, from the perspective of a single sample, as a noisy measurement of its scaled individual loss, with noise originating from stochastic batch composition. It is formally shown that the EMA mechanism functions as a first-order low-pass filter, attenuating high-frequency batch composition noise. This yields a score approximating the smoothed and persistent contribution of the individual sample to the loss, providing a theoretical grounding for BLS as a proxy for sample importance. BLS demonstrates remarkable code integration simplicity (\textbf{three-line injection}) and readily adapts existing per-sample loss-based methods (\textbf{one-line proxy}). Its effectiveness is demonstrated by enhancing two such methods to losslessly prune \textbf{20\%-50\%} of samples across \textit{14 datasets}, \textit{11 tasks} and \textit{18 models}, highlighting its utility and broad applicability, especially for complex scenarios where per-sample loss is difficult to access. Code is available at https://github.com/mrazhou/BLS.

Bingyu Li, Haocheng Dong, Da Zhang et al.

Open-Vocabulary Remote Sensing Image Segmentation (OVRSIS), an emerging task that adapts Open-Vocabulary Segmentation (OVS) to the remote sensing (RS) domain, remains underexplored due to the absence of a unified evaluation benchmark and the domain gap between natural and RS images. To bridge these gaps, we first establish a standardized OVRSIS benchmark (\textbf{OVRSISBench}) based on widely-used RS segmentation datasets, enabling consistent evaluation across methods. Using this benchmark, we comprehensively evaluate several representative OVS/OVRSIS models and reveal their limitations when directly applied to remote sensing scenarios. Building on these insights, we propose \textbf{RSKT-Seg}, a novel open-vocabulary segmentation framework tailored for remote sensing. RSKT-Seg integrates three key components: (1) a Multi-Directional Cost Map Aggregation (RS-CMA) module that captures rotation-invariant visual cues by computing vision-language cosine similarities across multiple directions; (2) an Efficient Cost Map Fusion (RS-Fusion) transformer, which jointly models spatial and semantic dependencies with a lightweight dimensionality reduction strategy; and (3) a Remote Sensing Knowledge Transfer (RS-Transfer) module that injects pre-trained knowledge and facilitates domain adaptation via enhanced upsampling. Extensive experiments on the benchmark show that RSKT-Seg consistently outperforms strong OVS baselines by +3.8 mIoU and +5.9 mACC, while achieving 2x faster inference through efficient aggregation. Our code is \href{https://github.com/LiBingyu01/RSKT-Seg}{\textcolor{blue}{here}}.

Junyu Gao, Da Zhang, Qiyu Wang et al.

Crowd localization targets on predicting each instance precise location within an image. Current advanced methods propose the pixel-wise binary classification to tackle the congested prediction, in which the pixel-level thresholds binarize the prediction confidence of being the pedestrian head. Since the crowd scenes suffer from extremely varying contents, counts and scales, the confidence-threshold learner is fragile and under-generalized encountering domain knowledge shift. Moreover, at the most time, the target domain is agnostic in training. Hence, it is imperative to exploit how to enhance the generalization of confidence-threshold locator to the latent target domain. In this paper, we propose a Dynamic Proxy Domain (DPD) method to generalize the learner under domain shift. Concretely, based on the theoretical analysis to the generalization error risk upper bound on the latent target domain to a binary classifier, we propose to introduce a generated proxy domain to facilitate generalization. Then, based on the theory, we design a DPD algorithm which is composed by a training paradigm and proxy domain generator to enhance the domain generalization of the confidence-threshold learner. Besides, we conduct our method on five kinds of domain shift scenarios, demonstrating the effectiveness on generalizing the crowd localization. Our code will be available at https://github.com/zhangda1018/DPD.

Qing Zhou, Shiyu Zhang, Yuyu Jia et al.

Chain-of-thought (CoT) reasoning has significantly improved the performance of large multimodal models in language-guided segmentation, yet its prohibitive computational cost, stemming from generating verbose rationales, limits real-world applicability. We introduce WISE (Wisdom from Internal Self-Exploration), a novel paradigm for efficient reasoning guided by the principle of \textit{thinking twice -- once for learning, once for speed}. WISE trains a model to generate a structured sequence: a concise rationale, the final answer, and then a detailed explanation. By placing the concise rationale first, our method leverages autoregressive conditioning to enforce that the concise rationale acts as a sufficient summary for generating the detailed explanation. This structure is reinforced by a self-distillation objective that jointly rewards semantic fidelity and conciseness, compelling the model to internalize its detailed reasoning into a compact form. At inference, the detailed explanation is omitted. To address the resulting conditional distribution shift, our inference strategy, WISE-S, employs a simple prompting technique that injects a brevity-focused instruction into the user's query. This final adjustment facilitates the robust activation of the learned concise policy, unlocking the full benefits of our framework. Extensive experiments show that WISE-S achieves state-of-the-art zero-shot performance on the ReasonSeg benchmark with 58.3 cIoU, while reducing the average reasoning length by nearly \textbf{5$\times$} -- from 112 to just 23 tokens. Code is available at \href{https://github.com/mrazhou/WISE}{WISE}.

Da Zhang, Chenggang Rong, Bingyu Li et al.

Large vision-language models (VLMs) have achieved remarkable success in natural scene understanding, yet their application to underwater environments remains largely unexplored. Underwater imagery presents unique challenges including severe light attenuation, color distortion, and suspended particle scattering, while requiring specialized knowledge of marine ecosystems and organism taxonomy. To bridge this gap, we introduce UWBench, a comprehensive benchmark specifically designed for underwater vision-language understanding. UWBench comprises 15,003 high-resolution underwater images captured across diverse aquatic environments, encompassing oceans, coral reefs, and deep-sea habitats. Each image is enriched with human-verified annotations including 15,281 object referring expressions that precisely describe marine organisms and underwater structures, and 124,983 question-answer pairs covering diverse reasoning capabilities from object recognition to ecological relationship understanding. The dataset captures rich variations in visibility, lighting conditions, and water turbidity, providing a realistic testbed for model evaluation. Based on UWBench, we establish three comprehensive benchmarks: detailed image captioning for generating ecologically informed scene descriptions, visual grounding for precise localization of marine organisms, and visual question answering for multimodal reasoning about underwater environments. Extensive experiments on state-of-the-art VLMs demonstrate that underwater understanding remains challenging, with substantial room for improvement. Our benchmark provides essential resources for advancing vision-language research in underwater contexts and supporting applications in marine science, ecological monitoring, and autonomous underwater exploration. Our code and benchmark will be available.

Junyu Gao, Liangliang Zhao, Xuelong Li

Object counting is a hot topic in computer vision, which aims to estimate the number of objects in a given image. However, most methods only count objects of a single category for an image, which cannot be applied to scenes that need to count objects with multiple categories simultaneously, especially in aerial scenes. To this end, this paper introduces a Multi-category Object Counting (MOC) task to estimate the numbers of different objects (cars, buildings, ships, etc.) in an aerial image. Considering the absence of a dataset for this task, a large-scale Dataset (NWPU-MOC) is collected, consisting of 3,416 scenes with a resolution of 1024 $\times$ 1024 pixels, and well-annotated using 14 fine-grained object categories. Besides, each scene contains RGB and Near Infrared (NIR) images, of which the NIR spectrum can provide richer characterization information compared with only the RGB spectrum. Based on NWPU-MOC, the paper presents a multi-spectrum, multi-category object counting framework, which employs a dual-attention module to fuse the features of RGB and NIR and subsequently regress multi-channel density maps corresponding to each object category. In addition, to modeling the dependency between different channels in the density map with each object category, a spatial contrast loss is designed as a penalty for overlapping predictions at the same spatial position. Experimental results demonstrate that the proposed method achieves state-of-the-art performance compared with some mainstream counting algorithms. The dataset, code and models are publicly available at https://github.com/lyongo/NWPU-MOC.

Junyu Gao, Mengyuan Chen, Changsheng Xu

We target at the task of weakly-supervised action localization (WSAL), where only video-level action labels are available during model training. Despite the recent progress, existing methods mainly embrace a localization-by-classification paradigm and overlook the fruitful fine-grained temporal distinctions between video sequences, thus suffering from severe ambiguity in classification learning and classification-to-localization adaption. This paper argues that learning by contextually comparing sequence-to-sequence distinctions offers an essential inductive bias in WSAL and helps identify coherent action instances. Specifically, under a differentiable dynamic programming formulation, two complementary contrastive objectives are designed, including Fine-grained Sequence Distance (FSD) contrasting and Longest Common Subsequence (LCS) contrasting, where the first one considers the relations of various action/background proposals by using match, insert, and delete operators and the second one mines the longest common subsequences between two videos. Both contrasting modules can enhance each other and jointly enjoy the merits of discriminative action-background separation and alleviated task gap between classification and localization. Extensive experiments show that our method achieves state-of-the-art performance on two popular benchmarks. Our code is available at https://github.com/MengyuanChen21/CVPR2022-FTCL.

Junyu Gao, Tao Han, Qi Wang et al.

Accurately locating each head's position in the crowd scenes is a crucial task in the field of crowd analysis. However, traditional density-based methods only predict coarse prediction, and segmentation/detection-based methods cannot handle extremely dense scenes and large-range scale-variations crowds. To this end, we propose an end-to-end and straightforward framework for crowd localization, named Independent Instance Map segmentation (IIM). Different from density maps and boxes regression, each instance in IIM is non-overlapped. By segmenting crowds into independent connected components, the positions and the crowd counts (the centers and the number of components, respectively) are obtained. Furthermore, to improve the segmentation quality for different density regions, we present a differentiable Binarization Module (BM) to output structured instance maps. BM brings two advantages into localization models: 1) adaptively learn a threshold map for different images to detect each instance more accurately; 2) directly train the model using loss on binary predictions and labels. Extensive experiments verify the proposed method is effective and outperforms the-state-of-the-art methods on the five popular crowd datasets. Significantly, IIM improves F1-measure by 10.4% on the NWPU-Crowd Localization task. The source code and pre-trained models will be released at https://github.com/taohan10200/IIM.

Tao Han, Junyu Gao, Yuan Yuan et al.

Unlabeled data learning has attracted considerable attention recently. However, it is still elusive to extract the expected high-level semantic feature with mere unsupervised learning. In the meantime, semi-supervised learning (SSL) demonstrates a promising future in leveraging few samples. In this paper, we combine both to propose an Unsupervised Semantic Aggregation and Deformable Template Matching (USADTM) framework for SSL, which strives to improve the classification performance with few labeled data and then reduce the cost in data annotating. Specifically, unsupervised semantic aggregation based on Triplet Mutual Information (T-MI) loss is explored to generate semantic labels for unlabeled data. Then the semantic labels are aligned to the actual class by the supervision of labeled data. Furthermore, a feature pool that stores the labeled samples is dynamically updated to assign proxy labels for unlabeled data, which are used as targets for cross-entropy minimization. Extensive experiments and analysis across four standard semi-supervised learning benchmarks validate that USADTM achieves top performance (e.g., 90.46$\%$ accuracy on CIFAR-10 with 40 labels and 95.20$\%$ accuracy with 250 labels). The code is released at https://github.com/taohan10200/USADTM.

Qi Wang, Tao Han, Junyu Gao et al.

Cross-domain crowd counting (CDCC) is a hot topic due to its importance in public safety. The purpose of CDCC is to alleviate the domain shift between the source and target domain. Recently, typical methods attempt to extract domain-invariant features via image translation and adversarial learning. When it comes to specific tasks, we find that the domain shifts are reflected on model parameters' differences. To describe the domain gap directly at the parameter-level, we propose a Neuron Linear Transformation (NLT) method, exploiting domain factor and bias weights to learn the domain shift. Specifically, for a specific neuron of a source model, NLT exploits few labeled target data to learn domain shift parameters. Finally, the target neuron is generated via a linear transformation. Extensive experiments and analysis on six real-world datasets validate that NLT achieves top performance compared with other domain adaptation methods. An ablation study also shows that the NLT is robust and more effective than supervised and fine-tune training. Code is available at: \url{https://github.com/taohan10200/NLT}.

Guangshuai Gao, Junyu Gao, Qingjie Liu et al.

Accurately estimating the number of objects in a single image is a challenging yet meaningful task and has been applied in many applications such as urban planning and public safety. In the various object counting tasks, crowd counting is particularly prominent due to its specific significance to social security and development. Fortunately, the development of the techniques for crowd counting can be generalized to other related fields such as vehicle counting and environment survey, if without taking their characteristics into account. Therefore, many researchers are devoting to crowd counting, and many excellent works of literature and works have spurted out. In these works, they are must be helpful for the development of crowd counting. However, the question we should consider is why they are effective for this task. Limited by the cost of time and energy, we cannot analyze all the algorithms. In this paper, we have surveyed over 220 works to comprehensively and systematically study the crowd counting models, mainly CNN-based density map estimation methods. Finally, according to the evaluation metrics, we select the top three performers on their crowd counting datasets and analyze their merits and drawbacks. Through our analysis, we expect to make reasonable inference and prediction for the future development of crowd counting, and meanwhile, it can also provide feasible solutions for the problem of object counting in other fields. We provide the density maps and prediction results of some mainstream algorithm in the validation set of NWPU dataset for comparison and testing. Meanwhile, density map generation and evaluation tools are also provided. All the codes and evaluation results are made publicly available at https://github.com/gaoguangshuai/survey-for-crowd-counting.