Wentong Li

Xue Yang, Gefan Zhang, Wentong Li et al.

Oriented object detection emerges in many applications from aerial images to autonomous driving, while many existing detection benchmarks are annotated with horizontal bounding box only which is also less costive than fine-grained rotated box, leading to a gap between the readily available training corpus and the rising demand for oriented object detection. This paper proposes a simple yet effective oriented object detection approach called H2RBox merely using horizontal box annotation for weakly-supervised training, which closes the above gap and shows competitive performance even against those trained with rotated boxes. The cores of our method are weakly- and self-supervised learning, which predicts the angle of the object by learning the consistency of two different views. To our best knowledge, H2RBox is the first horizontal box annotation-based oriented object detector. Compared to an alternative i.e. horizontal box-supervised instance segmentation with our post adaption to oriented object detection, our approach is not susceptible to the prediction quality of mask and can perform more robustly in complex scenes containing a large number of dense objects and outliers. Experimental results show that H2RBox has significant performance and speed advantages over horizontal box-supervised instance segmentation methods, as well as lower memory requirements. While compared to rotated box-supervised oriented object detectors, our method shows very close performance and speed. The source code is available at PyTorch-based \href{https://github.com/yangxue0827/h2rbox-mmrotate}{MMRotate} and Jittor-based \href{https://github.com/yangxue0827/h2rbox-jittor}{JDet}.

Wentong Li, Wenyu Liu, Jianke Zhu et al.

In contrast to the fully supervised methods using pixel-wise mask labels, box-supervised instance segmentation takes advantage of the simple box annotations, which has recently attracted a lot of research attentions. In this paper, we propose a novel single-shot box-supervised instance segmentation approach, which integrates the classical level set model with deep neural network delicately. Specifically, our proposed method iteratively learns a series of level sets through a continuous Chan-Vese energy-based function in an end-to-end fashion. A simple mask supervised SOLOv2 model is adapted to predict the instance-aware mask map as the level set for each instance. Both the input image and its deep features are employed as the input data to evolve the level set curves, where a box projection function is employed to obtain the initial boundary. By minimizing the fully differentiable energy function, the level set for each instance is iteratively optimized within its corresponding bounding box annotation. The experimental results on four challenging benchmarks demonstrate the leading performance of our proposed approach to robust instance segmentation in various scenarios. The code is available at: https://github.com/LiWentomng/boxlevelset.

Wentong Li, Wenyu Liu, Jianke Zhu et al.

In contrast to fully supervised methods using pixel-wise mask labels, box-supervised instance segmentation takes advantage of simple box annotations, which has recently attracted increasing research attention. This paper presents a novel single-shot instance segmentation approach, namely Box2Mask, which integrates the classical level-set evolution model into deep neural network learning to achieve accurate mask prediction with only bounding box supervision. Specifically, both the input image and its deep features are employed to evolve the level-set curves implicitly, and a local consistency module based on a pixel affinity kernel is used to mine the local context and spatial relations. Two types of single-stage frameworks, i.e., CNN-based and transformer-based frameworks, are developed to empower the level-set evolution for box-supervised instance segmentation, and each framework consists of three essential components: instance-aware decoder, box-level matching assignment and level-set evolution. By minimizing the level-set energy function, the mask map of each instance can be iteratively optimized within its bounding box annotation. The experimental results on five challenging testbeds, covering general scenes, remote sensing, medical and scene text images, demonstrate the outstanding performance of our proposed Box2Mask approach for box-supervised instance segmentation. In particular, with the Swin-Transformer large backbone, our Box2Mask obtains 42.4% mask AP on COCO, which is on par with the recently developed fully mask-supervised methods. The code is available at: https://github.com/LiWentomng/boxlevelset.

Wentong Li, Yuqian Yuan, Song Wang et al.

Weakly-supervised image segmentation has recently attracted increasing research attentions, aiming to avoid the expensive pixel-wise labeling. In this paper, we present an effective method, namely Point2Mask, to achieve high-quality panoptic prediction using only a single random point annotation per target for training. Specifically, we formulate the panoptic pseudo-mask generation as an Optimal Transport (OT) problem, where each ground-truth (gt) point label and pixel sample are defined as the label supplier and consumer, respectively. The transportation cost is calculated by the introduced task-oriented maps, which focus on the category-wise and instance-wise differences among the various thing and stuff targets. Furthermore, a centroid-based scheme is proposed to set the accurate unit number for each gt point supplier. Hence, the pseudo-mask generation is converted into finding the optimal transport plan at a globally minimal transportation cost, which can be solved via the Sinkhorn-Knopp Iteration. Experimental results on Pascal VOC and COCO demonstrate the promising performance of our proposed Point2Mask approach to point-supervised panoptic segmentation. Source code is available at: https://github.com/LiWentomng/Point2Mask.

Song Wang, Wentong Li, Wenyu Liu et al.

Semantic map construction under bird's-eye view (BEV) plays an essential role in autonomous driving. In contrast to camera image, LiDAR provides the accurate 3D observations to project the captured 3D features onto BEV space inherently. However, the vanilla LiDAR-based BEV feature often contains many indefinite noises, where the spatial features have little texture and semantic cues. In this paper, we propose an effective LiDAR-based method to build semantic map. Specifically, we introduce a BEV feature pyramid decoder that learns the robust multi-scale BEV features for semantic map construction, which greatly boosts the accuracy of the LiDAR-based method. To mitigate the defects caused by lacking semantic cues in LiDAR data, we present an online Camera-to-LiDAR distillation scheme to facilitate the semantic learning from image to point cloud. Our distillation scheme consists of feature-level and logit-level distillation to absorb the semantic information from camera in BEV. The experimental results on challenging nuScenes dataset demonstrate the efficacy of our proposed LiDAR2Map on semantic map construction, which significantly outperforms the previous LiDAR-based methods over 27.9% mIoU and even performs better than the state-of-the-art camera-based approaches. Source code is available at: https://github.com/songw-zju/LiDAR2Map.

Wentong Li, Yuqian Yuan, Jian Liu et al.

The visual projector serves as an essential bridge between the visual encoder and the Large Language Model (LLM) in a Multimodal LLM (MLLM). Typically, MLLMs adopt a simple MLP to preserve all visual contexts via one-to-one transformation. However, the visual tokens are redundant and can be considerably increased when dealing with high-resolution images, impairing the efficiency of MLLMs significantly. Some recent works have introduced resampler or abstractor to reduce the number of resulting visual tokens. Unfortunately, they fail to capture finer details and undermine the visual reasoning capabilities of MLLMs. In this work, we propose a novel visual projector, which adopts a coarse-to-fine scheme to inject the enriched characteristics to generate the condensed visual tokens. In specific, we first interpolate the visual features as a low-resolution point query, providing the overall visual representation as the foundation. Then, we introduce a region-to-point injection module that utilizes high-resolution, multi-level region-based cues as fine-grained reference keys and values, allowing them to be fully absorbed within the corresponding local context region. This step effectively updates the coarse point query, transforming it into an enriched one for the subsequent LLM reasoning. Extensive experiments demonstrate that our approach compresses the visual tokens by 75%~89%, while achieves comparable or even better performance across diverse benchmarks with significantly higher efficiency. The source codes can be found at https://github.com/CircleRadon/TokenPacker.

Wenyu Liu, Wentong Li, Jianke Zhu et al.

Semantic segmentation on driving-scene images is vital for autonomous driving. Although encouraging performance has been achieved on daytime images, the performance on nighttime images are less satisfactory due to the insufficient exposure and the lack of labeled data. To address these issues, we present an add-on module called dual image-adaptive learnable filters (DIAL-Filters) to improve the semantic segmentation in nighttime driving conditions, aiming at exploiting the intrinsic features of driving-scene images under different illuminations. DIAL-Filters consist of two parts, including an image-adaptive processing module (IAPM) and a learnable guided filter (LGF). With DIAL-Filters, we design both unsupervised and supervised frameworks for nighttime driving-scene segmentation, which can be trained in an end-to-end manner. Specifically, the IAPM module consists of a small convolutional neural network with a set of differentiable image filters, where each image can be adaptively enhanced for better segmentation with respect to the different illuminations. The LGF is employed to enhance the output of segmentation network to get the final segmentation result. The DIAL-Filters are light-weight and efficient and they can be readily applied for both daytime and nighttime images. Our experiments show that DAIL-Filters can significantly improve the supervised segmentation performance on ACDC_Night and NightCity datasets, while it demonstrates the state-of-the-art performance on unsupervised nighttime semantic segmentation on Dark Zurich and Nighttime Driving testbeds.

Yue Feng, Jingjing Li, Qijia Lu et al.

The rapid advancement of AI-driven video generation has transformed content creation, while simultaneously increasing the risk of misinformation through localized manipulations in long-form videos. Existing video forensic methods predominantly operate on short, independent clips, and thus fail to capture realistic scenarios where AI-generated content is sparsely embedded within otherwise authentic footage. To bridge this gap, we formulate the task of Temporal AI-Generated Segment Localization and Explanation, which targets authenticity detection, temporal localization, and interpretable analysis of manipulated segments in untrimmed long videos. We further introduce TASLE, a large-scale benchmark comprising 12,472 untrimmed videos with diverse manipulation patterns and rich annotation signals, including temporal boundaries, authenticity labels, and segment-level rationales. In addition, we propose MSLoc, a coarse-to-fine forensic baseline that combines a boundary-sensitive proposal generation module for efficient long-video scanning with an MLLM-based refinement module for precise boundary localization and interpretable reasoning. Experiments validate the effectiveness of the proposed baseline, highlighting the importance of segment-level explainable forensics for long-form AI-generated video analysis. Our dataset and code are publicly available at https://debby-0527.github.io/TASLE.

Qijun Gan, Wentong Li, Jinwei Ren et al.

Reconstructing high-fidelity hand models with intricate textures plays a crucial role in enhancing human-object interaction and advancing real-world applications. Despite the state-of-the-art methods excelling in texture generation and image rendering, they often face challenges in accurately capturing geometric details. Learning-based approaches usually offer better robustness and faster inference, which tend to produce smoother results and require substantial amounts of training data. To address these issues, we present a novel fine-grained multi-view hand mesh reconstruction method that leverages inverse rendering to restore hand poses and intricate details. Firstly, our approach predicts a parametric hand mesh model through Graph Convolutional Networks (GCN) based method from multi-view images. We further introduce a novel Hand Albedo and Mesh (HAM) optimization module to refine both the hand mesh and textures, which is capable of preserving the mesh topology. In addition, we suggest an effective mesh-based neural rendering scheme to simultaneously generate photo-realistic image and optimize mesh geometry by fusing the pre-trained rendering network with vertex features. We conduct the comprehensive experiments on InterHand2.6M, DeepHandMesh and dataset collected by ourself, whose promising results show that our proposed approach outperforms the state-of-the-art methods on both reconstruction accuracy and rendering quality. Code and dataset are publicly available at https://github.com/agnJason/FMHR.

Zihao Xin, Wentong Li, Yixuan Jiang et al.

Vision-and-Language Navigation (VLN) requires an embodied agent to ground complex natural-language instructions into long-horizon navigation in unseen environments. While Vision-Language Models (VLMs) offer strong 2D semantic understanding, current VLN systems remain constrained by limited spatial perception, 2D-3D representation mismatch, and monocular scale ambiguity. In this paper, we propose AgentVLN, a novel and efficient embodied navigation framework that can be deployed on edge computing platforms. We formulate VLN as a Partially Observable Semi-Markov Decision Process (POSMDP) and introduce a VLM-as-Brain paradigm that decouples high-level semantic reasoning from perception and planning via a plug-and-play skill library. To resolve multi-level representation inconsistency, we design a cross-space representation mapping that projects perception-layer 3D topological waypoints into the image plane, yielding pixel-aligned visual prompts for the VLM. Building on this bridge, we integrate a context-aware self-correction and active exploration strategy to recover from occlusions and suppress error accumulation over long trajectories. To further address the spatial ambiguity of instructions in unstructured environments, we propose a Query-Driven Perceptual Chain-of-Thought (QD-PCoT) scheme, enabling the agent with the metacognitive ability to actively seek geometric depth information. Finally, we construct AgentVLN-Instruct, a large-scale instruction-tuning dataset with dynamic stage routing conditioned on target visibility. Extensive experiments show that AgentVLN consistently outperforms prior state-of-the-art methods (SOTA) on long-horizon VLN benchmarks, offering a practical paradigm for lightweight deployment of next-generation embodied navigation models. Code: https://github.com/Allenxinn/AgentVLN.

Hanxun Yu, Wentong Li, Xuan Qu et al.

Multimodal large language models (MLLMs) suffer from high computational costs due to excessive visual tokens, particularly in high-resolution and video-based scenarios. Existing token reduction methods typically focus on isolated pipeline components and often neglect textual alignment, leading to performance degradation. In this paper, we propose VisionTrim, a unified framework for training-free MLLM acceleration, integrating two effective plug-and-play modules: 1) the Dominant Vision Token Selection (DVTS) module, which preserves essential visual tokens via a global-local view, and 2) the Text-Guided Vision Complement (TGVC) module, which facilitates context-aware token merging guided by textual cues. Extensive experiments across diverse image and video multimodal benchmarks demonstrate the performance superiority of our VisionTrim, advancing practical MLLM deployment in real-world applications. The code is available at: https://github.com/hanxunyu/VisionTrim.

Yuqian Yuan, Wenqiao Zhang, Juekai Lin et al.

Large Multimodal Models (LMMs) have achieved remarkable progress in general-purpose vision--language understanding, yet they remain limited in tasks requiring precise object-level grounding, fine-grained spatial reasoning, and controllable visual manipulation. In particular, existing systems often struggle to identify the correct instance, preserve object identity across interactions, and localize or modify designated regions with high precision. Object-centric vision provides a principled framework for addressing these challenges by promoting explicit representations and operations over visual entities, thereby extending multimodal systems from global scene understanding to object-level understanding, segmentation, editing, and generation. This paper presents a comprehensive review of recent advances at the convergence of LMMs and object-centric vision. We organize the literature into four major themes: object-centric visual understanding, object-centric referring segmentation, object-centric visual editing, and object-centric visual generation. We further summarize the key modeling paradigms, learning strategies, and evaluation protocols that support these capabilities. Finally, we discuss open challenges and future directions, including robust instance permanence, fine-grained spatial control, consistent multi-step interaction, unified cross-task modeling, and reliable benchmarking under distribution shift. We hope this paper provides a structured perspective on the development of scalable, precise, and trustworthy object-centric multimodal systems.

Yuqian Yuan, Wentong Li, Jian Liu et al.

Multimodal large language models (MLLMs) have recently achieved impressive general-purpose vision-language capabilities through visual instruction tuning. However, current MLLMs primarily focus on image-level or box-level understanding, falling short in achieving fine-grained vision-language alignment at pixel level. Besides, the lack of mask-based instruction data limits their advancements. In this paper, we propose Osprey, a mask-text instruction tuning approach, to extend MLLMs by incorporating fine-grained mask regions into language instruction, aiming at achieving pixel-wise visual understanding. To achieve this goal, we first meticulously curate a mask-based region-text dataset with 724K samples, and then design a vision-language model by injecting pixel-level representation into LLM. Specifically, Osprey adopts a convolutional CLIP backbone as the vision encoder and employs a mask-aware visual extractor to extract precise visual mask features from high resolution input. Experimental results demonstrate Osprey's superiority in various region understanding tasks, showcasing its new capability for pixel-level instruction tuning. In particular, Osprey can be integrated with Segment Anything Model (SAM) seamlessly to obtain multi-granularity semantics. The source code, dataset and demo can be found at https://github.com/CircleRadon/Osprey.

Yuqian Yuan, Wentong Li, Zhaocheng Li et al.

In this paper, we introduce InstructSAM, a unified and streamlined framework designed for multi-instance segmentation under arbitrary instructions. We formulates instruction-driven instance segmentation as a set-structured query prediction problem and propose an explicit reasoning-to-instance query interface that elegantly bridges a vision-language model (VLM) and SAM3. Specifically, a bank of learnable instance queries is injected into the VLM and contextualized with instruction and visual information, enabling each query to serve as an instance-aware slot. A hybrid-attention mechanism further promotes interaction among these queries, visual tokens, and instruction tokens, improving instance enumeration and reducing duplicate predictions. The resulting LLM-conditioned queries are projected into SAM3's detector query space to drive accurate multi-instance segmentation in a single forward pass. This design equips SAM3 with high-level instruction understanding, compositional reasoning, and instance-level set prediction without modifying its core architecture. To support training and evaluation, we further construct Inst2Seg, a high-quality and large-scale instruction-based instance segmentation dataset and benchmark that couples free-form instructions with instance-level masks. Extensive experiments show that only 2B-scale InstructSAM achieves strong results across complex instruction-driven and phrase-level referring segmentation benchmarks, outperforming prior end-to-end methods and SAM3's agentic pipeline while enabling efficient single-pass multi-instance prediction.

Wentong Li, Yuqian Yuan, Song Wang et al.

Weakly-supervised segmentation with label-efficient sparse annotations has attracted increasing research attention to reduce the cost of laborious pixel-wise labeling process, while the pairwise affinity modeling techniques play an essential role in this task. Most of the existing approaches focus on using the local appearance kernel to model the neighboring pairwise potentials. However, such a local operation fails to capture the long-range dependencies and ignores the topology of objects. In this work, we formulate the affinity modeling as an affinity propagation process, and propose a local and a global pairwise affinity terms to generate accurate soft pseudo labels. An efficient algorithm is also developed to reduce significantly the computational cost. The proposed approach can be conveniently plugged into existing segmentation networks. Experiments on three typical label-efficient segmentation tasks, i.e. box-supervised instance segmentation, point/scribble-supervised semantic segmentation and CLIP-guided semantic segmentation, demonstrate the superior performance of the proposed approach.

Xiaolu Liu, Song Wang, Wentong Li et al.

Currently, high-definition (HD) map construction leans towards a lightweight online generation tendency, which aims to preserve timely and reliable road scene information. However, map elements contain strong shape priors. Subtle and sparse annotations make current detection-based frameworks ambiguous in locating relevant feature scopes and cause the loss of detailed structures in prediction. To alleviate these problems, we propose MGMap, a mask-guided approach that effectively highlights the informative regions and achieves precise map element localization by introducing the learned masks. Specifically, MGMap employs learned masks based on the enhanced multi-scale BEV features from two perspectives. At the instance level, we propose the Mask-activated instance (MAI) decoder, which incorporates global instance and structural information into instance queries by the activation of instance masks. At the point level, a novel position-guided mask patch refinement (PG-MPR) module is designed to refine point locations from a finer-grained perspective, enabling the extraction of point-specific patch information. Compared to the baselines, our proposed MGMap achieves a notable improvement of around 10 mAP for different input modalities. Extensive experiments also demonstrate that our approach showcases strong robustness and generalization capabilities. Our code can be found at https://github.com/xiaolul2/MGMap.

Song Wang, Jiawei Yu, Wentong Li et al.

Semantic scene completion, also known as semantic occupancy prediction, can provide dense geometric and semantic information for autonomous vehicles, which attracts the increasing attention of both academia and industry. Unfortunately, existing methods usually formulate this task as a voxel-wise classification problem and treat each voxel equally in 3D space during training. As the hard voxels have not been paid enough attention, the performance in some challenging regions is limited. The 3D dense space typically contains a large number of empty voxels, which are easy to learn but require amounts of computation due to handling all the voxels uniformly for the existing models. Furthermore, the voxels in the boundary region are more challenging to differentiate than those in the interior. In this paper, we propose HASSC approach to train the semantic scene completion model with hardness-aware design. The global hardness from the network optimization process is defined for dynamical hard voxel selection. Then, the local hardness with geometric anisotropy is adopted for voxel-wise refinement. Besides, self-distillation strategy is introduced to make training process stable and consistent. Extensive experiments show that our HASSC scheme can effectively promote the accuracy of the baseline model without incurring the extra inference cost. Source code is available at: https://github.com/songw-zju/HASSC.

Song Wang, Zhongdao Wang, Jiawei Yu et al.

Vision-centric semantic occupancy prediction plays a crucial role in autonomous driving, which requires accurate and reliable predictions from low-cost sensors. Although having notably narrowed the accuracy gap with LiDAR, there is still few research effort to explore the reliability in predicting semantic occupancy from camera. In this paper, we conduct a comprehensive evaluation of existing semantic occupancy prediction models from a reliability perspective for the first time. Despite the gradual alignment of camera-based models with LiDAR in term of accuracy, a significant reliability gap persists. To addresses this concern, we propose ReliOcc, a method designed to enhance the reliability of camera-based occupancy networks. ReliOcc provides a plug-and-play scheme for existing models, which integrates hybrid uncertainty from individual voxels with sampling-based noise and relative voxels through mix-up learning. Besides, an uncertainty-aware calibration strategy is devised to further enhance model reliability in offline mode. Extensive experiments under various settings demonstrate that ReliOcc significantly enhances model reliability while maintaining the accuracy of both geometric and semantic predictions. Importantly, our proposed approach exhibits robustness to sensor failures and out of domain noises during inference.

Hanxun Yu, Wentong Li, Song Wang et al.

Despite encouraging progress in 3D scene understanding, it remains challenging to develop an effective Large Multi-modal Model (LMM) that is capable of understanding and reasoning in complex 3D environments. Most previous methods typically encode 3D point and 2D image features separately, neglecting interactions between 2D semantics and 3D object properties, as well as the spatial relationships within the 3D environment. This limitation not only hinders comprehensive representations of 3D scene, but also compromises training and inference efficiency. To address these challenges, we propose a unified Instance-aware 3D Large Multi-modal Model (Inst3D-LMM) to deal with multiple 3D scene understanding tasks simultaneously. To obtain the fine-grained instance-level visual tokens, we first introduce a novel Multi-view Cross-Modal Fusion (MCMF) module to inject the multi-view 2D semantics into their corresponding 3D geometric features. For scene-level relation-aware tokens, we further present a 3D Instance Spatial Relation (3D-ISR) module to capture the intricate pairwise spatial relationships among objects. Additionally, we perform end-to-end multi-task instruction tuning simultaneously without the subsequent task-specific fine-tuning. Extensive experiments demonstrate that our approach outperforms the state-of-the-art methods across 3D scene understanding, reasoning and grounding tasks. Source code is available at https://github.com/hanxunyu/Inst3D-LMM

Wentong Li, Zhiyuan Qi, Zichen Zhao et al.

Pre-trained vision foundation models (VFMs) provide strong semantic representations, yet their patch-level features are inherently coarse, limiting their effectiveness on tasks requiring fine-grained localization, dense prediction, and point-wise correspondence. In this work, we revisit feature upsampling for VFMs from the perspective of \textbf{\textit{inverse problem}} and propose Weighted Reverse Convolution (WRC), a spatially adaptive inverse operator for densifying high-level visual descriptors. Specifically, we formulate feature upsampling as a weighted Tikhonov-regularized least-squares problem, where spatially varying weights modulate both data fidelity and prior strength at each spatial location. This allows WRC to adapt the reconstruction to spatially varying feature characteristics, thereby preserving critical structures while mitigating over-smoothing. Moreover, WRC retains an efficient, fully differentiable closed-form FFT solution, making it a practical drop-in upsampling operator. Integrated into a lightweight self-supervised densification framework, WRC consistently improves dense feature quality across various downstream benchmarks, including segmentation, depth estimation, video object segmentation, object discovery, and keypoint correspondence, while maintaining high computational efficiency.

Song Wang, Lingdong Kong, Xiaolu Liu et al.

The rapid advancement of autonomous systems, including self-driving vehicles and drones, has intensified the need to forge true Spatial Intelligence from multi-modal onboard sensor data. While foundation models excel in single-modal contexts, integrating their capabilities across diverse sensors like cameras and LiDAR to create a unified understanding remains a formidable challenge. This paper presents a comprehensive framework for multi-modal pre-training, identifying the core set of techniques driving progress toward this goal. We dissect the interplay between foundational sensor characteristics and learning strategies, evaluating the role of platform-specific datasets in enabling these advancements. Our central contribution is the formulation of a unified taxonomy for pre-training paradigms: ranging from single-modality baselines to sophisticated unified frameworks that learn holistic representations for advanced tasks like 3D object detection and semantic occupancy prediction. Furthermore, we investigate the integration of textual inputs and occupancy representations to facilitate open-world perception and planning. Finally, we identify critical bottlenecks, such as computational efficiency and model scalability, and propose a roadmap toward general-purpose multi-modal foundation models capable of achieving robust Spatial Intelligence for real-world deployment.

Bin Wang, Wentong Li, Wenqian Wang et al.

Recently, large-scale pre-trained vision-language models (e.g., CLIP), have garnered significant attention thanks to their powerful representative capabilities. This inspires researchers in transferring the knowledge from these large pre-trained models to other task-specific models, e.g., Video Action Recognition (VAR) models, via particularly leveraging side networks to enhance the efficiency of parameter-efficient fine-tuning (PEFT). However, current transferring approaches in VAR tend to directly transfer the frozen knowledge from large pre-trained models to action recognition networks with minimal cost, instead of exploiting the temporal modeling capabilities of the action recognition models themselves. Therefore, in this paper, we propose a novel memory-efficient Temporal Difference Side Network (TDS-CLIP) to balance knowledge transferring and temporal modeling, avoiding backpropagation in frozen parameter models. Specifically, we introduce a Temporal Difference Adapter (TD-Adapter), which can effectively capture local temporal differences in motion features to strengthen the model's global temporal modeling capabilities. Furthermore, we designed a Side Motion Enhancement Adapter (SME-Adapter) to guide the proposed side network in efficiently learning the rich motion information in videos, thereby improving the side network's ability to capture and learn motion information. Extensive experiments are conducted on three benchmark datasets, including Something-Something V1&V2, and Kinetics-400. Experimental results show that our method achieves competitive performance in video action recognition tasks.

Yuqian Yuan, Ronghao Dang, Long Li et al.

The emergence of multimodal large language models (MLLMs) has driven breakthroughs in egocentric vision applications. These applications necessitate persistent, context-aware understanding of objects, as users interact with tools in dynamic and cluttered environments. However, existing embodied benchmarks primarily focus on static scene exploration, emphasizing object's appearance and spatial attributes while neglecting the assessment of dynamic changes arising from users' interactions. To address this gap, we introduce EOC-Bench, an innovative benchmark designed to systematically evaluate object-centric embodied cognition in dynamic egocentric scenarios. Specially, EOC-Bench features 3,277 meticulously annotated QA pairs categorized into three temporal categories: Past, Present, and Future, covering 11 fine-grained evaluation dimensions and 3 visual object referencing types. To ensure thorough assessment, we develop a mixed-format human-in-the-loop annotation framework with four types of questions and design a novel multi-scale temporal accuracy metric for open-ended temporal evaluation. Based on EOC-Bench, we conduct comprehensive evaluations of various proprietary, open-source, and object-level MLLMs. EOC-Bench serves as a crucial tool for advancing the embodied object cognitive capabilities of MLLMs, establishing a robust foundation for developing reliable core models for embodied systems.

Song Wang, Xiaolu Liu, Lingdong Kong et al.

Self-supervised representation learning for point cloud has demonstrated effectiveness in improving pre-trained model performance across diverse tasks. However, as pre-trained models grow in complexity, fully fine-tuning them for downstream applications demands substantial computational and storage resources. Parameter-efficient fine-tuning (PEFT) methods offer a promising solution to mitigate these resource requirements, yet most current approaches rely on complex adapter and prompt mechanisms that increase tunable parameters. In this paper, we propose PointLoRA, a simple yet effective method that combines low-rank adaptation (LoRA) with multi-scale token selection to efficiently fine-tune point cloud models. Our approach embeds LoRA layers within the most parameter-intensive components of point cloud transformers, reducing the need for tunable parameters while enhancing global feature capture. Additionally, multi-scale token selection extracts critical local information to serve as prompts for downstream fine-tuning, effectively complementing the global context captured by LoRA. The experimental results across various pre-trained models and three challenging public datasets demonstrate that our approach achieves competitive performance with only 3.43% of the trainable parameters, making it highly effective for resource-constrained applications. Source code is available at: https://github.com/songw-zju/PointLoRA.

Zihao Xin, Wentong Li, Yixuan Jiang et al.

Vision-and-Language Navigation (VLN) requires agents to follow long-horizon instructions and navigate complex 3D environments. However, existing approaches face two major challenges: constructing an effective long-term memory bank and overcoming the compounding errors problem. To address these issues, we propose DecoVLN, an effective framework designed for robust streaming perception and closed-loop control in long-horizon navigation. First, we formulate long-term memory construction as an optimization problem and introduce adaptive refinement mechanism that selects frames from a historical candidate pool by iteratively optimizing a unified scoring function. This function jointly balances three key criteria: semantic relevance to the instruction, visual diversity from the selected memory, and temporal coverage of the historical trajectory. Second, to alleviate compounding errors, we introduce a state-action pair-level corrective finetuning strategy. By leveraging geodesic distance between states to precisely quantify deviation from the expert trajectory, the agent collects high-quality state-action pairs in the trusted region while filtering out the polluted data with low relevance. This improves both the efficiency and stability of error correction. Extensive experiments demonstrate the effectiveness of DecoVLN, and we have deployed it in real-world environments.

Yue Feng, Jinwei Hu, Qijia Lu et al.

We propose the Multi-modal Untrimmed Video Retrieval task, along with a new benchmark (MUVR) to advance video retrieval for long-video platforms. MUVR aims to retrieve untrimmed videos containing relevant segments using multi-modal queries. It has the following features: 1) Practical retrieval paradigm: MUVR supports video-centric multi-modal queries, expressing fine-grained retrieval needs through long text descriptions, video tag prompts, and mask prompts. It adopts a one-to-many retrieval paradigm and focuses on untrimmed videos, tailored for long-video platform applications. 2) Multi-level visual correspondence: To cover common video categories (e.g., news, travel, dance) and precisely define retrieval matching criteria, we construct multi-level visual correspondence based on core video content (e.g., news events, travel locations, dance moves) which users are interested in and want to retrieve. It covers six levels: copy, event, scene, instance, action, and others. 3) Comprehensive evaluation criteria: We develop 3 versions of MUVR (i.e., Base, Filter, QA). MUVR-Base/Filter evaluates retrieval models, while MUVR-QA assesses MLLMs in a question-answering format. We also propose a Reranking Score to evaluate the reranking ability of MLLMs. MUVR consists of 53K untrimmed videos from the video platform Bilibili, with 1,050 multi-modal queries and 84K matches. Extensive evaluations of 3 state-of-the-art video retrieval models, 6 image-based VLMs, and 10 MLLMs are conducted. MUVR reveals the limitations of retrieval methods in processing untrimmed videos and multi-modal queries, as well as MLLMs in multi-video understanding and reranking. Our code and benchmark is available at https://github.com/debby-0527/MUVR.

Xiaolu Liu, Ruizi Yang, Song Wang et al.

Reliable high-definition (HD) map construction is crucial for the driving safety of autonomous vehicles. Although recent studies demonstrate improved performance, their generalization capability across unfamiliar driving scenes remains unexplored. To tackle this issue, we propose UIGenMap, an uncertainty-instructed structure injection approach for generalizable HD map vectorization, which concerns the uncertainty resampling in statistical distribution and employs explicit instance features to reduce excessive reliance on training data. Specifically, we introduce the perspective-view (PV) detection branch to obtain explicit structural features, in which the uncertainty-aware decoder is designed to dynamically sample probability distributions considering the difference in scenes. With probabilistic embedding and selection, UI2DPrompt is proposed to construct PV-learnable prompts. These PV prompts are integrated into the map decoder by designed hybrid injection to compensate for neglected instance structures. To ensure real-time inference, a lightweight Mimic Query Distillation is designed to learn from PV prompts, which can serve as an efficient alternative to the flow of PV branches. Extensive experiments on challenging geographically disjoint (geo-based) data splits demonstrate that our UIGenMap achieves superior performance, with +5.7 mAP improvement on the nuScenes dataset. Source code will be available at https://github.com/xiaolul2/UIGenMap.

Wentong Li, Yijie Chen, Wenyu Liu et al.

Box-supervised instance segmentation has recently attracted lots of research efforts while little attention is received in aerial image domain. In contrast to the general object collections, aerial objects have large intra-class variances and inter-class similarity with complex background. Moreover, there are many tiny objects in the high-resolution satellite images. This makes the recent pairwise affinity modeling method inevitably to involve the noisy supervision with the inferior results. To tackle these problems, we propose a novel aerial instance segmentation approach, which drives the network to learn a series of level set functions for the aerial objects with only box annotations in an end-to-end fashion. Instead of learning the pairwise affinity, the level set method with the carefully designed energy functions treats the object segmentation as curve evolution, which is able to accurately recover the object's boundaries and prevent the interference from the indistinguishable background and similar objects. The experimental results demonstrate that the proposed approach outperforms the state-of-the-art box-supervised instance segmentation methods. The source code is available at https://github.com/LiWentomng/boxlevelset.

Wentong Li, Yijie Chen, Kaixuan Hu et al.

In contrast to the generic object, aerial targets are often non-axis aligned with arbitrary orientations having the cluttered surroundings. Unlike the mainstreamed approaches regressing the bounding box orientations, this paper proposes an effective adaptive points learning approach to aerial object detection by taking advantage of the adaptive points representation, which is able to capture the geometric information of the arbitrary-oriented instances. To this end, three oriented conversion functions are presented to facilitate the classification and localization with accurate orientation. Moreover, we propose an effective quality assessment and sample assignment scheme for adaptive points learning toward choosing the representative oriented reppoints samples during training, which is able to capture the non-axis aligned features from adjacent objects or background noises. A spatial constraint is introduced to penalize the outlier points for roust adaptive learning. Experimental results on four challenging aerial datasets including DOTA, HRSC2016, UCAS-AOD and DIOR-R, demonstrate the efficacy of our proposed approach. The source code is availabel at: https://github.com/LiWentomng/OrientedRepPoints.

Yuqian Yuan, Hang Zhang, Wentong Li et al. · pku

Video Large Language Models (Video LLMs) have recently exhibited remarkable capabilities in general video understanding. However, they mainly focus on holistic comprehension and struggle with capturing fine-grained spatial and temporal details. Besides, the lack of high-quality object-level video instruction data and a comprehensive benchmark further hinders their advancements. To tackle these challenges, we introduce the VideoRefer Suite to empower Video LLM for finer-level spatial-temporal video understanding, i.e., enabling perception and reasoning on any objects throughout the video. Specially, we thoroughly develop VideoRefer Suite across three essential aspects: dataset, model, and benchmark. Firstly, we introduce a multi-agent data engine to meticulously curate a large-scale, high-quality object-level video instruction dataset, termed VideoRefer-700K. Next, we present the VideoRefer model, which equips a versatile spatial-temporal object encoder to capture precise regional and sequential representations. Finally, we meticulously create a VideoRefer-Bench to comprehensively assess the spatial-temporal understanding capability of a Video LLM, evaluating it across various aspects. Extensive experiments and analyses demonstrate that our VideoRefer model not only achieves promising performance on video referring benchmarks but also facilitates general video understanding capabilities.

Jinhong Wang, Jian Liu, Dongqi Tang et al.

This paper shows that the autoregressive model is an effective and scalable monocular depth estimator. Our idea is simple: We tackle the monocular depth estimation (MDE) task with an autoregressive prediction paradigm, based on two core designs. First, our depth autoregressive model (DAR) treats the depth map of different resolutions as a set of tokens, and conducts the low-to-high resolution autoregressive objective with a patch-wise casual mask. Second, our DAR recursively discretizes the entire depth range into more compact intervals, and attains the coarse-to-fine granularity autoregressive objective in an ordinal-regression manner. By coupling these two autoregressive objectives, our DAR establishes new state-of-the-art (SOTA) on KITTI and NYU Depth v2 by clear margins. Further, our scalable approach allows us to scale the model up to 2.0B and achieve the best RMSE of 1.799 on the KITTI dataset (5% improvement) compared to 1.896 by the current SOTA (Depth Anything). DAR further showcases zero-shot generalization ability on unseen datasets. These results suggest that DAR yields superior performance with an autoregressive prediction paradigm, providing a promising approach to equip modern autoregressive large models (e.g., GPT-4o) with depth estimation capabilities.

Bin Wang, Ruotong Hu, Wenqian Wang et al.

Visual and textual soft prompt tuning can effectively improve the adaptability of Vision-Language Models (VLMs) in downstream tasks. However, fine-tuning on video tasks impairs the model's generalization ability to unseen classes. Existing methods attempt to mitigate this forgetting effect by regularizing the gap between hand-crafted prompts and soft prompts, but this also weakens the learning ability of soft prompts. To address this challenge, we propose a plug-and-play coupling prompt learning framework to optimize the generalization performance of V-L models in video tasks, with the core motivation of mitigating semantic space narrowing during fine-tuning by introducing an externally supervised prompt. Specifically, for textual prompts, we introduce pre-trained prompts from other datasets as hard prompt tokens. These are concatenated with soft prompt tokens and coupled via a learnable mapping layer. This competitive prompting approach prevents the semantic space from overfitting to supervised categories. In addition, we introduce a set of well-designed irrelevant video sets and negative prompts as generic attribute anchors to maintain the generic relevance of the attributes in the pre-trained semantic space, thus preserving the generalization ability. Experiments on video tasks demonstrate that our method significantly outperforms state-of-the-art prompt tuning approaches across generalization benchmarks, particularly on base-to-new class prediction.

Yuhang Qian, Haiyan Chen, Wentong Li et al.

Camouflage Images Generation (CIG) is an emerging research area that focuses on synthesizing images in which objects are harmoniously blended and exhibit high visual consistency with their surroundings. Existing methods perform CIG by either fusing objects into specific backgrounds or outpainting the surroundings via foreground object-guided diffusion. However, they often fail to obtain natural results because they overlook the logical relationship between camouflaged objects and background environments. To address this issue, we propose CT-CIG, a Controllable Text-guided Camouflage Images Generation method that produces realistic and logically plausible camouflage images. Leveraging Large Visual Language Models (VLM), we design a Camouflage-Revealing Dialogue Mechanism (CRDM) to annotate existing camouflage datasets with high-quality text prompts. Subsequently, the constructed image-prompt pairs are utilized to finetune Stable Diffusion, incorporating a lightweight controller to guide the location and shape of camouflaged objects for enhanced camouflage scene fitness. Moreover, we design a Frequency Interaction Refinement Module (FIRM) to capture high-frequency texture features, facilitating the learning of complex camouflage patterns. Extensive experiments, including CLIPScore evaluation and camouflage effectiveness assessment, demonstrate the semantic alignment of our generated text prompts and CT-CIG's ability to produce photorealistic camouflage images.

Yuqian Yuan, Wenqiao Zhang, Xin Li et al.

Multimodal large language models (MLLMs) have demonstrated strong general-purpose capabilities in open-world visual comprehension. However, most existing MLLMs primarily focus on holistic, scene-level understanding, often overlooking the need for fine-grained, object-centric reasoning. In this paper, we present PixelRefer, a unified region-level MLLM framework that enables advanced fine-grained understanding over user-specified regions across both images and videos. Motivated by the observation that LLM attention predominantly focuses on object-level tokens, we propose a Scale-Adaptive Object Tokenizer (SAOT) to generate compact and semantically rich object representations from free-form regions. Our analysis reveals that global visual tokens contribute mainly in early LLM layers, inspiring the design of PixelRefer-Lite, an efficient variant that employs an Object-Centric Infusion module to pre-fuse global context into object tokens. This yields a lightweight Object-Only Framework that substantially reduces computational cost while maintaining high semantic fidelity. To facilitate fine-grained instruction tuning, we curate PixelRefer-2.2M, a high-quality object-centric instruction dataset. Extensive experiments across a range of benchmarks validate that PixelRefer achieves leading performance with fewer training samples, while PixelRefer-Lite offers competitive accuracy with notable gains in efficiency.

Teng Zhang, Ziqian Fan, Mingxin Liu et al.

Driven by the growing need for Oriented Object Detection (OOD), learning from point annotations under a weakly-supervised framework has emerged as a promising alternative to costly and laborious manual labeling. In this paper, we discuss two deficiencies in existing point-supervised methods: inefficient utilization and poor quality of pseudo labels. Therefore, we present Point2RBox-v3. At the core are two principles: 1) Progressive Label Assignment (PLA). It dynamically estimates instance sizes in a coarse yet intelligent manner at different stages of the training process, enabling the use of label assignment methods. 2) Prior-Guided Dynamic Mask Loss (PGDM-Loss). It is an enhancement of the Voronoi Watershed Loss from Point2RBox-v2, which overcomes the shortcomings of Watershed in its poor performance in sparse scenes and SAM's poor performance in dense scenes. To our knowledge, Point2RBox-v3 is the first model to employ dynamic pseudo labels for label assignment, and it creatively complements the advantages of SAM model with the watershed algorithm, which achieves excellent performance in both sparse and dense scenes. Our solution gives competitive performance, especially in scenarios with large variations in object size or sparse object occurrences: 66.09%/56.86%/41.28%/46.40%/19.60%/45.96% on DOTA-v1.0/DOTA-v1.5/DOTA-v2.0/DIOR/STAR/RSAR.

Xinglong Yang, Quan Feng, Zhongying Pan et al.

The effectiveness of Multimodal Chain-of-Thought (MCoT) prompting is often limited by the use of randomly or manually selected examples. These examples fail to account for both model-specific knowledge distributions and the intrinsic complexity of the tasks, resulting in suboptimal and unstable model performance. To address this, we propose a novel framework inspired by the pedagogical principle of "tailored teaching with balanced difficulty". We reframe prompt selection as a prompt curriculum design problem: constructing a well ordered set of training examples that align with the model's current capabilities. Our approach integrates two complementary signals: (1) model-perceived difficulty, quantified through prediction disagreement in an active learning setup, capturing what the model itself finds challenging; and (2) intrinsic sample complexity, which measures the inherent difficulty of each question-image pair independently of any model. By jointly analyzing these signals, we develop a difficulty-balanced sampling strategy that ensures the selected prompt examples are diverse across both dimensions. Extensive experiments conducted on five challenging benchmarks and multiple popular Multimodal Large Language Models (MLLMs) demonstrate that our method yields substantial and consistent improvements and greatly reduces performance discrepancies caused by random sampling, providing a principled and robust approach for enhancing multimodal reasoning.

Jinhong Wang, Shuo Tong, Jian liu et al.

Despite the remarkable progress of multimodal large language models (MLLMs), they continue to face challenges in achieving competitive performance on ordinal regression (OR; a.k.a. ordinal classification). To address this issue, this paper presents OrderChain, a novel and general prompting paradigm that improves the ordinal understanding ability of MLLMs by specificity and commonality modeling. Specifically, our OrderChain consists of a set of task-aware prompts to facilitate the specificity modeling of diverse OR tasks and a new range optimization Chain-of-Thought (RO-CoT), which learns a commonality way of thinking about OR tasks by uniformly decomposing them into multiple small-range optimization subtasks. Further, we propose a category recursive division (CRD) method to generate instruction candidate category prompts to support RO-CoT automatic optimization. Comprehensive experiments show that LLaVA model with our OrderChain improves baseline LLaVA significantly on diverse OR datasets, e.g., from 47.5\% to 93.2\% accuracy on the Adience dataset for age estimation, and from 30.0\% to 85.7\% accuracy on the Diabetic Retinopathy dataset. Notably, LLaVA with our OrderChain also remarkably outperforms state-of-the-art methods by 27% on accuracy and 0.24 on MAE on the Adience dataset. To our best knowledge, our OrderChain is the first work that augments MLLMs for OR tasks, and the effectiveness is witnessed across a spectrum of OR datasets. Project Page: https://order-chain.github.io/.

Hantang Liu, Wentong Li, Jianke Zhu

Effectively parsing the facade is essential to 3D building reconstruction, which is an important computer vision problem with a large amount of applications in high precision map for navigation, computer aided design, and city generation for digital entertainments. To this end, the key is how to obtain the shape grammars from 2D images accurately and efficiently. Although enjoying the merits of promising results on the semantic parsing, deep learning methods cannot directly make use of the architectural rules, which play an important role for man-made structures. In this paper, we present a novel translational symmetry-based approach to improving the deep neural networks. Our method employs deep learning models as the base parser, and a module taking advantage of translational symmetry is used to refine the initial parsing results. In contrast to conventional semantic segmentation or bounding box prediction, we propose a novel scheme to fuse segmentation with anchor-free detection in a single stage network, which enables the efficient training and better convergence. After parsing the facades into shape grammars, we employ an off-the-shelf rendering engine like Blender to reconstruct the realistic high-quality 3D models using procedural modeling. We conduct experiments on three public datasets, where our proposed approach outperforms the state-of-the-art methods. In addition, we have illustrated the 3D building models built from 2D facade images.

Jian Xu, Kuang-chih Lee, Wentong Li et al.

In targeted online advertising, advertisers look for maximizing campaign performance under delivery constraint within budget schedule. Most of the advertisers typically prefer to impose the delivery constraint to spend budget smoothly over the time in order to reach a wider range of audiences and have a sustainable impact. Since lots of impressions are traded through public auctions for online advertising today, the liquidity makes price elasticity and bid landscape between demand and supply change quite dynamically. Therefore, it is challenging to perform smooth pacing control and maximize campaign performance simultaneously. In this paper, we propose a smart pacing approach in which the delivery pace of each campaign is learned from both offline and online data to achieve smooth delivery and optimal performance goals. The implementation of the proposed approach in a real DSP system is also presented. Experimental evaluations on both real online ad campaigns and offline simulations show that our approach can effectively improve campaign performance and achieve delivery goals.