Meng Wei

Jin Ye, Haoyu Wang, Ziyan Huang et al.

Tumor lesion segmentation is one of the most important tasks in medical image analysis. In clinical practice, Fluorodeoxyglucose Positron-Emission Tomography~(FDG-PET) is a widely used technique to identify and quantify metabolically active tumors. However, since FDG-PET scans only provide metabolic information, healthy tissue or benign disease with irregular glucose consumption may be mistaken for cancer. To handle this challenge, PET is commonly combined with Computed Tomography~(CT), with the CT used to obtain the anatomic structure of the patient. The combination of PET-based metabolic and CT-based anatomic information can contribute to better tumor segmentation results. %Computed tomography~(CT) is a popular modality to illustrate the anatomic structure of the patient. The combination of PET and CT is promising to handle this challenge by utilizing metabolic and anatomic information. In this paper, we explore the potential of U-Net for lesion segmentation in whole-body FDG-PET/CT scans from three aspects, including network architecture, data preprocessing, and data augmentation. The experimental results demonstrate that the vanilla U-Net with proper input shape can achieve satisfactory performance. Specifically, our method achieves first place in both preliminary and final leaderboards of the autoPET 2022 challenge. Our code is available at https://github.com/Yejin0111/autoPET2022_Blackbean.

Zijian Zhou, Oluwatosin Alabi, Meng Wei et al.

In this paper, we propose a novel text promptable surgical instrument segmentation approach to overcome challenges associated with diversity and differentiation of surgical instruments in minimally invasive surgeries. We redefine the task as text promptable, thereby enabling a more nuanced comprehension of surgical instruments and adaptability to new instrument types. Inspired by recent advancements in vision-language models, we leverage pretrained image and text encoders as our model backbone and design a text promptable mask decoder consisting of attention- and convolution-based prompting schemes for surgical instrument segmentation prediction. Our model leverages multiple text prompts for each surgical instrument through a new mixture of prompts mechanism, resulting in enhanced segmentation performance. Additionally, we introduce a hard instrument area reinforcement module to improve image feature comprehension and segmentation precision. Extensive experiments on several surgical instrument segmentation datasets demonstrate our model's superior performance and promising generalization capability. To our knowledge, this is the first implementation of a promptable approach to surgical instrument segmentation, offering significant potential for practical application in the field of robotic-assisted surgery. Code is available at https://github.com/franciszzj/TP-SIS.

Yuncheng Yang, Meng Wei, Junjun He et al.

Transfer learning is a critical technique in training deep neural networks for the challenging medical image segmentation task that requires enormous resources. With the abundance of medical image data, many research institutions release models trained on various datasets that can form a huge pool of candidate source models to choose from. Hence, it's vital to estimate the source models' transferability (i.e., the ability to generalize across different downstream tasks) for proper and efficient model reuse. To make up for its deficiency when applying transfer learning to medical image segmentation, in this paper, we therefore propose a new Transferability Estimation (TE) method. We first analyze the drawbacks of using the existing TE algorithms for medical image segmentation and then design a source-free TE framework that considers both class consistency and feature variety for better estimation. Extensive experiments show that our method surpasses all current algorithms for transferability estimation in medical image segmentation. Code is available at https://github.com/EndoluminalSurgicalVision-IMR/CCFV

Meng Wei, Xiaoyu Yue, Wenwei Zhang et al.

Segmenting and recognizing diverse object parts is a crucial ability in applications spanning various computer vision and robotic tasks. While significant progress has been made in object-level Open-Vocabulary Semantic Segmentation (OVSS), i.e., segmenting objects with arbitrary text, the corresponding part-level research poses additional challenges. Firstly, part segmentation inherently involves intricate boundaries, while limited annotated data compounds the challenge. Secondly, part segmentation introduces an open granularity challenge due to the diverse and often ambiguous definitions of parts in the open world. Furthermore, the large-scale vision and language models, which play a key role in the open vocabulary setting, struggle to recognize parts as effectively as objects. To comprehensively investigate and tackle these challenges, we propose an Open-Vocabulary Part Segmentation (OV-PARTS) benchmark. OV-PARTS includes refined versions of two publicly available datasets: Pascal-Part-116 and ADE20K-Part-234. And it covers three specific tasks: Generalized Zero-Shot Part Segmentation, Cross-Dataset Part Segmentation, and Few-Shot Part Segmentation, providing insights into analogical reasoning, open granularity and few-shot adapting abilities of models. Moreover, we analyze and adapt two prevailing paradigms of existing object-level OVSS methods for OV-PARTS. Extensive experimental analysis is conducted to inspire future research in leveraging foundational models for OV-PARTS. The code and dataset are available at https://github.com/OpenRobotLab/OV_PARTS.

Team Seedance, De Chen, Liyang Chen et al. · gatech

Seedance 2.0 is a new native multi-modal audio-video generation model, officially released in China in early February 2026. Compared with its predecessors, Seedance 1.0 and 1.5 Pro, Seedance 2.0 adopts a unified, highly efficient, and large-scale architecture for multi-modal audio-video joint generation. This allows it to support four input modalities: text, image, audio, and video, by integrating one of the most comprehensive suites of multi-modal content reference and editing capabilities available in the industry to date. It delivers substantial, well-rounded improvements across all key sub-dimensions of video and audio generation. In both expert evaluations and public user tests, the model has demonstrated performance on par with the leading levels in the field. Seedance 2.0 supports direct generation of audio-video content with durations ranging from 4 to 15 seconds, with native output resolutions of 480p and 720p. For multi-modal inputs as reference, its current open platform supports up to 3 video clips, 9 images, and 3 audio clips. In addition, we provide Seedance 2.0 Fast version, an accelerated variant of Seedance 2.0 designed to boost generation speed for low-latency scenarios. Seedance 2.0 has delivered significant improvements to its foundational generation capabilities and multi-modal generation performance, bringing an enhanced creative experience for end users.

Yanda Meng, Joshua Bridge, Meng Wei et al.

This paper proposes an adaptive auxiliary task learning based approach for object counting problems. Unlike existing auxiliary task learning based methods, we develop an attention-enhanced adaptively shared backbone network to enable both task-shared and task-tailored features learning in an end-to-end manner. The network seamlessly combines standard Convolution Neural Network (CNN) and Graph Convolution Network (GCN) for feature extraction and feature reasoning among different domains of tasks. Our approach gains enriched contextual information by iteratively and hierarchically fusing the features across different task branches of the adaptive CNN backbone. The whole framework pays special attention to the objects' spatial locations and varied density levels, informed by object (or crowd) segmentation and density level segmentation auxiliary tasks. In particular, thanks to the proposed dilated contrastive density loss function, our network benefits from individual and regional context supervision in terms of pixel-independent and pixel-dependent feature learning mechanisms, along with strengthened robustness. Experiments on seven challenging multi-domain datasets demonstrate that our method achieves superior performance to the state-of-the-art auxiliary task learning based counting methods. Our code is made publicly available at: https://github.com/smallmax00/Counting_With_Adaptive_Auxiliary

Meng Wei, Charlie Budd, Luis C. Garcia-Peraza-Herrera et al.

Surgical instrument segmentation is recognised as a key enabler in providing advanced surgical assistance and improving computer-assisted interventions. In this work, we propose SegMatch, a semi-supervised learning method to reduce the need for expensive annotation for laparoscopic and robotic surgical images. SegMatch builds on FixMatch, a widespread semi supervised classification pipeline combining consistency regularization and pseudo-labelling, and adapts it for the purpose of segmentation. In our proposed SegMatch, the unlabelled images are first weakly augmented and fed to the segmentation model to generate pseudo-labels. In parallel, images are fed to a strong augmentation branch and consistency between the branches is used as an unsupervised loss. To increase the relevance of our strong augmentations, we depart from using only handcrafted augmentations and introduce a trainable adversarial augmentation strategy. Our FixMatch adaptation for segmentation tasks further includes carefully considering the equivariance and invariance properties of the augmentation functions we rely on. For binary segmentation tasks, our algorithm was evaluated on the MICCAI Instrument Segmentation Challenge datasets, Robust-MIS 2019 and EndoVis 2017. For multi-class segmentation tasks, we relied on the recent CholecInstanceSeg dataset. Our results show that SegMatch outperforms fully-supervised approaches by incorporating unlabelled data, and surpasses a range of state-of-the-art semi-supervised models across different labelled to unlabelled data ratios.

Meng Wei, Long Chen, Wei Ji et al.

Video Visual Relation Detection (VidVRD) aims to detect visual relationship triplets in videos using spatial bounding boxes and temporal boundaries. Existing VidVRD methods can be broadly categorized into bottom-up and top-down paradigms, depending on their approach to classifying relations. Bottom-up methods follow a clip-based approach where they classify relations of short clip tubelet pairs and then merge them into long video relations. On the other hand, top-down methods directly classify long video tubelet pairs. While recent video-based methods utilizing video tubelets have shown promising results, we argue that the effective modeling of spatial and temporal context plays a more significant role than the choice between clip tubelets and video tubelets. This motivates us to revisit the clip-based paradigm and explore the key success factors in VidVRD. In this paper, we propose a Hierarchical Context Model (HCM) that enriches the object-based spatial context and relation-based temporal context based on clips. We demonstrate that using clip tubelets can achieve superior performance compared to most video-based methods. Additionally, using clip tubelets offers more flexibility in model designs and helps alleviate the limitations associated with video tubelets, such as the challenging long-term object tracking problem and the loss of temporal information in long-term tubelet feature compression. Extensive experiments conducted on two challenging VidVRD benchmarks validate that our HCM achieves a new state-of-the-art performance, highlighting the effectiveness of incorporating advanced spatial and temporal context modeling within the clip-based paradigm.

Xiaoyu Yue, Lei Bai, Meng Wei et al.

Masked AutoEncoder (MAE) has revolutionized the field of self-supervised learning with its simple yet effective masking and reconstruction strategies. However, despite achieving state-of-the-art performance across various downstream vision tasks, the underlying mechanisms that drive MAE's efficacy are less well-explored compared to the canonical contrastive learning paradigm. In this paper, we first propose a local perspective to explicitly extract a local contrastive form from MAE's reconstructive objective at the patch level. And then we introduce a new empirical framework, called Local Contrastive MAE (LC-MAE), to analyze both reconstructive and contrastive aspects of MAE. LC-MAE reveals that MAE learns invariance to random masking and ensures distribution consistency between the learned token embeddings and the original images. Furthermore, we dissect the contribution of the decoder and random masking to MAE's success, revealing both the decoder's learning mechanism and the dual role of random masking as data augmentation and effective receptive field restriction. Our experimental analysis sheds light on the intricacies of MAE and summarizes some useful design methodologies, which can inspire more powerful visual self-supervised methods.

Ruozhen He, Meng Wei, Ziyan Yang et al.

Multi-shot video generation extends single-shot generation to coherent visual narratives, yet maintaining consistent characters, objects, and locations across shots remains a challenge over long sequences. Existing evaluations typically use independently generated prompt sets with limited entity coverage and simple consistency metrics, making standardized comparison difficult. We introduce EntityBench, a benchmark of 140 episodes (2,491 shots) derived from real narrative media, with explicit per-shot entity schedules tracking characters, objects, and locations simultaneously across easy / medium / hard tiers of up to 50 shots, 13 cross-shot characters, 8 cross-shot locations, 22 cross-shot objects, and recurrence gaps spanning up to 48 shots. It is paired with a three-pillar evaluation suite that disentangles intra-shot quality, prompt-following alignment, and cross-shot consistency, with a fidelity gate that admits only accurate entity appearances into cross-shot scoring. As a baseline, we propose EntityMem, a memory-augmented generation system that stores verified per-entity visual references in a persistent memory bank before generation begins. Experiments show that cross-shot entity consistency degrades sharply with recurrence distance in existing methods, and that explicit per-entity memory yields the highest character fidelity (Cohen's d = +2.33) and presence among methods evaluated. Code and data are available at https://github.com/Catherine-R-He/EntityBench/.

Cheng Zhang, Boying Li, Meng Wei et al.

Transformers have emerged as a universal backbone across 3D perception, video generation, and world models for autonomous driving and embodied AI, where understanding camera geometry is essential for grounding visual observations in three-dimensional space. However, existing camera encoding methods often rely on simplified pinhole assumptions, restricting generalization across the diverse intrinsics and lens distortions in real-world cameras. We introduce Relative Ray Encoding, a geometry-consistent representation that unifies complete camera information, including 6-DoF poses, intrinsics, and lens distortions. To evaluate its capability under diverse controllability demands, we adopt camera-controlled text-to-video generation as a testbed task. Within this setting, we further identify pitch and roll as two components effective for Absolute Orientation Encoding, enabling full control over the initial camera orientation. Together, these designs form UCPE (Unified Camera Positional Encoding), which integrates into a pretrained video Diffusion Transformer through a lightweight spatial attention adapter, adding less than 1% trainable parameters while achieving state-of-the-art camera controllability and visual fidelity. To facilitate systematic training and evaluation, we construct a large video dataset covering a wide range of camera motions and lens types. Extensive experiments validate the effectiveness of UCPE in camera-controllable video generation and highlight its potential as a general camera representation for Transformers across future multi-view, video, and 3D tasks. Code will be available at https://github.com/chengzhag/UCPE.

Meng Wei, Yong Zhou, Zhongnian Li et al.

Complementary-label learning (CLL) is widely used in weakly supervised classification, but it faces a significant challenge in real-world datasets when confronted with class-imbalanced training samples. In such scenarios, the number of samples in one class is considerably lower than in other classes, which consequently leads to a decline in the accuracy of predictions. Unfortunately, existing CLL approaches have not investigate this problem. To alleviate this challenge, we propose a novel problem setting that enables learning from class-imbalanced complementary labels for multi-class classification. To tackle this problem, we propose a novel CLL approach called Weighted Complementary-Label Learning (WCLL). The proposed method models a weighted empirical risk minimization loss by utilizing the class-imbalanced complementary labels, which is also applicable to multi-class imbalanced training samples. Furthermore, we derive an estimation error bound to provide theoretical assurance. To evaluate our approach, we conduct extensive experiments on several widely-used benchmark datasets and a real-world dataset, and compare our method with existing state-of-the-art methods. The proposed approach shows significant improvement in these datasets, even in the case of multiple class-imbalanced scenarios. Notably, the proposed method not only utilizes complementary labels to train a classifier but also solves the problem of class imbalance.

Meng Wei, Kun Yuan, Shi Li et al.

Enabling intuitive, language-driven interaction with surgical scenes is a critical step toward intelligent operating rooms and autonomous surgical robotic assistance. However, the task of referring segmentation, localizing surgical instruments based on natural language descriptions, remains underexplored in surgical videos, with existing approaches struggling to generalize due to reliance on static visual cues and predefined instrument names. In this work, we introduce SurgRef, a novel motion-guided framework that grounds free-form language expressions in instrument motion, capturing how tools move and interact across time, rather than what they look like. This allows models to understand and segment instruments even under occlusion, ambiguity, or unfamiliar terminology. To train and evaluate SurgRef, we present Ref-IMotion, a diverse, multi-institutional video dataset with dense spatiotemporal masks and rich motion-centric expressions. SurgRef achieves state-of-the-art accuracy and generalization across surgical procedures, setting a new benchmark for robust, language-driven surgical video segmentation.

Yuwei Guo, Ceyuan Yang, Hao He et al.

Autoregressive video diffusion models hold promise for world simulation but are vulnerable to exposure bias arising from the train-test mismatch. While recent works address this via post-training, they typically rely on a bidirectional teacher model or online discriminator. To achieve an end-to-end solution, we introduce Resampling Forcing, a teacher-free framework that enables training autoregressive video models from scratch and at scale. Central to our approach is a self-resampling scheme that simulates inference-time model errors on history frames during training. Conditioned on these degraded histories, a sparse causal mask enforces temporal causality while enabling parallel training with frame-level diffusion loss. To facilitate efficient long-horizon generation, we further introduce history routing, a parameter-free mechanism that dynamically retrieves the top-k most relevant history frames for each query. Experiments demonstrate that our approach achieves performance comparable to distillation-based baselines while exhibiting superior temporal consistency on longer videos owing to native-length training.

Gheorghe Comanici, Eric Bieber, Mike Schaekermann et al. · amazon-science, baidu

In this report, we introduce the Gemini 2.X model family: Gemini 2.5 Pro and Gemini 2.5 Flash, as well as our earlier Gemini 2.0 Flash and Flash-Lite models. Gemini 2.5 Pro is our most capable model yet, achieving SoTA performance on frontier coding and reasoning benchmarks. In addition to its incredible coding and reasoning skills, Gemini 2.5 Pro is a thinking model that excels at multimodal understanding and it is now able to process up to 3 hours of video content. Its unique combination of long context, multimodal and reasoning capabilities can be combined to unlock new agentic workflows. Gemini 2.5 Flash provides excellent reasoning abilities at a fraction of the compute and latency requirements and Gemini 2.0 Flash and Flash-Lite provide high performance at low latency and cost. Taken together, the Gemini 2.X model generation spans the full Pareto frontier of model capability vs cost, allowing users to explore the boundaries of what is possible with complex agentic problem solving.

Meng Wei, Cheng Zhang, Jianmin Zheng et al.

Recent advances have equipped 3D Gaussian Splatting with texture parameterizations to capture spatially varying attributes, improving the performance of both appearance modeling and downstream tasks. However, the added texture parameters introduce significant memory efficiency challenges. Rather than proposing new texture formulations, we take a step back to examine the characteristics of existing textured Gaussian methods and identify two key limitations in common: (1) Textures are typically defined in canonical space, leading to inefficient sampling that wastes textures' capacity on low-contribution regions; and (2) texture parameterization is uniformly assigned across all Gaussians, regardless of their visual complexity, resulting in over-parameterization. In this work, we address these issues through two simple yet effective strategies: adaptive sampling based on the Gaussian density distribution and error-driven anisotropic parameterization that allocates texture resources according to rendering error. Our proposed ASAP Textured Gaussians, short for Adaptive Sampling and Anisotropic Parameterization, significantly improve the quality efficiency tradeoff, achieving high-fidelity rendering with far fewer texture parameters.

Oluwatosin Alabi, Meng Wei, Charlie Budd et al.

Understanding surgical instrument-tissue interactions requires not only identifying which instrument performs which action on which anatomical target, but also grounding these interactions spatially within the surgical scene. Existing surgical action triplet recognition methods are limited to learning from frame-level classification, failing to reliably link actions to specific instrument instances.Previous attempts at spatial grounding have primarily relied on class activation maps, which lack the precision and robustness required for detailed instrument-tissue interaction analysis.To address this gap, we propose grounding surgical action triplets with instrument instance segmentation, or triplet segmentation for short, a new unified task which produces spatially grounded <instrument, verb, target> outputs.We start by presenting CholecTriplet-Seg, a large-scale dataset containing over 30,000 annotated frames, linking instrument instance masks with action verb and anatomical target annotations, and establishing the first benchmark for strongly supervised, instance-level triplet grounding and evaluation.To learn triplet segmentation, we propose TargetFusionNet, a novel architecture that extends Mask2Former with a target-aware fusion mechanism to address the challenge of accurate anatomical target prediction by fusing weak anatomy priors with instrument instance queries.Evaluated across recognition, detection, and triplet segmentation metrics, TargetFusionNet consistently improves performance over existing baselines, demonstrating that strong instance supervision combined with weak target priors significantly enhances the accuracy and robustness of surgical action understanding.Triplet segmentation establishes a unified framework for spatially grounding surgical action triplets. The proposed benchmark and architecture pave the way for more interpretable, surgical scene understanding.

Meng Wei, Long Chen, Wei Ji et al.

Grounded Situation Recognition (GSR), i.e., recognizing the salient activity (or verb) category in an image (e.g., buying) and detecting all corresponding semantic roles (e.g., agent and goods), is an essential step towards "human-like" event understanding. Since each verb is associated with a specific set of semantic roles, all existing GSR methods resort to a two-stage framework: predicting the verb in the first stage and detecting the semantic roles in the second stage. However, there are obvious drawbacks in both stages: 1) The widely-used cross-entropy (XE) loss for object recognition is insufficient in verb classification due to the large intra-class variation and high inter-class similarity among daily activities. 2) All semantic roles are detected in an autoregressive manner, which fails to model the complex semantic relations between different roles. To this end, we propose a novel SituFormer for GSR which consists of a Coarse-to-Fine Verb Model (CFVM) and a Transformer-based Noun Model (TNM). CFVM is a two-step verb prediction model: a coarse-grained model trained with XE loss first proposes a set of verb candidates, and then a fine-grained model trained with triplet loss re-ranks these candidates with enhanced verb features (not only separable but also discriminative). TNM is a transformer-based semantic role detection model, which detects all roles parallelly. Owing to the global relation modeling ability and flexibility of the transformer decoder, TNM can fully explore the statistical dependency of the roles. Extensive validations on the challenging SWiG benchmark show that SituFormer achieves a new state-of-the-art performance with significant gains under various metrics. Code is available at https://github.com/kellyiss/SituFormer.

Xiaoyu Yue, Shuyang Sun, Zhanghui Kuang et al.

Transformers with powerful global relation modeling abilities have been introduced to fundamental computer vision tasks recently. As a typical example, the Vision Transformer (ViT) directly applies a pure transformer architecture on image classification, by simply splitting images into tokens with a fixed length, and employing transformers to learn relations between these tokens. However, such naive tokenization could destruct object structures, assign grids to uninterested regions such as background, and introduce interference signals. To mitigate the above issues, in this paper, we propose an iterative and progressive sampling strategy to locate discriminative regions. At each iteration, embeddings of the current sampling step are fed into a transformer encoder layer, and a group of sampling offsets is predicted to update the sampling locations for the next step. The progressive sampling is differentiable. When combined with the Vision Transformer, the obtained PS-ViT network can adaptively learn where to look. The proposed PS-ViT is both effective and efficient. When trained from scratch on ImageNet, PS-ViT performs 3.8% higher than the vanilla ViT in terms of top-1 accuracy with about $4\times$ fewer parameters and $10\times$ fewer FLOPs. Code is available at https://github.com/yuexy/PS-ViT.

Meng Wei, Zhongnian Li, Yong Zhou et al.

Annotating multi-class instances is a crucial task in the field of machine learning. Unfortunately, identifying the correct class label from a long sequence of candidate labels is time-consuming and laborious. To alleviate this problem, we design a novel labeling mechanism called stochastic label. In this setting, stochastic label includes two cases: 1) identify a correct class label from a small number of randomly given labels; 2) annotate the instance with None label when given labels do not contain correct class label. In this paper, we propose a novel suitable approach to learn from these stochastic labels. We obtain an unbiased estimator that utilizes less supervised information in stochastic labels to train a multi-class classifier. Additionally, it is theoretically justifiable by deriving the estimation error bound of the proposed method. Finally, we conduct extensive experiments on widely-used benchmark datasets to validate the superiority of our method by comparing it with existing state-of-the-art methods.

Team Seawead, Ceyuan Yang, Zhijie Lin et al.

This technical report presents a cost-efficient strategy for training a video generation foundation model. We present a mid-sized research model with approximately 7 billion parameters (7B) called Seaweed-7B trained from scratch using 665,000 H100 GPU hours. Despite being trained with moderate computational resources, Seaweed-7B demonstrates highly competitive performance compared to contemporary video generation models of much larger size. Design choices are especially crucial in a resource-constrained setting. This technical report highlights the key design decisions that enhance the performance of the medium-sized diffusion model. Empirically, we make two observations: (1) Seaweed-7B achieves performance comparable to, or even surpasses, larger models trained on substantially greater GPU resources, and (2) our model, which exhibits strong generalization ability, can be effectively adapted across a wide range of downstream applications either by lightweight fine-tuning or continue training. See the project page at https://seaweed.video/

Hao He, Ceyuan Yang, Shanchuan Lin et al.

This paper introduces CameraCtrl II, a framework that enables large-scale dynamic scene exploration through a camera-controlled video diffusion model. Previous camera-conditioned video generative models suffer from diminished video dynamics and limited range of viewpoints when generating videos with large camera movement. We take an approach that progressively expands the generation of dynamic scenes -- first enhancing dynamic content within individual video clip, then extending this capability to create seamless explorations across broad viewpoint ranges. Specifically, we construct a dataset featuring a large degree of dynamics with camera parameter annotations for training while designing a lightweight camera injection module and training scheme to preserve dynamics of the pretrained models. Building on these improved single-clip techniques, we enable extended scene exploration by allowing users to iteratively specify camera trajectories for generating coherent video sequences. Experiments across diverse scenarios demonstrate that CameraCtrl Ii enables camera-controlled dynamic scene synthesis with substantially wider spatial exploration than previous approaches.

Jianyi Wang, Zhijie Lin, Meng Wei et al.

Video restoration poses non-trivial challenges in maintaining fidelity while recovering temporally consistent details from unknown degradations in the wild. Despite recent advances in diffusion-based restoration, these methods often face limitations in generation capability and sampling efficiency. In this work, we present SeedVR, a diffusion transformer designed to handle real-world video restoration with arbitrary length and resolution. The core design of SeedVR lies in the shifted window attention that facilitates effective restoration on long video sequences. SeedVR further supports variable-sized windows near the boundary of both spatial and temporal dimensions, overcoming the resolution constraints of traditional window attention. Equipped with contemporary practices, including causal video autoencoder, mixed image and video training, and progressive training, SeedVR achieves highly-competitive performance on both synthetic and real-world benchmarks, as well as AI-generated videos. Extensive experiments demonstrate SeedVR's superiority over existing methods for generic video restoration.

Meng Wei, Chenyang Wan, Xiqian Yu et al.

Vision-and-Language Navigation (VLN) in real-world settings requires agents to process continuous visual streams and generate actions with low latency grounded in language instructions. While Video-based Large Language Models (Video-LLMs) have driven recent progress, current VLN methods based on Video-LLM often face trade-offs among fine-grained visual understanding, long-term context modeling and computational efficiency. We introduce StreamVLN, a streaming VLN framework that employs a hybrid slow-fast context modeling strategy to support multi-modal reasoning over interleaved vision, language and action inputs. The fast-streaming dialogue context facilitates responsive action generation through a sliding-window of active dialogues, while the slow-updating memory context compresses historical visual states using a 3D-aware token pruning strategy. With this slow-fast design, StreamVLN achieves coherent multi-turn dialogue through efficient KV cache reuse, supporting long video streams with bounded context size and inference cost. Experiments on VLN-CE benchmarks demonstrate state-of-the-art performance with stable low latency, ensuring robustness and efficiency in real-world deployment. The project page is: \href{https://streamvln.github.io/}{https://streamvln.github.io/}.

Zijian Zhou, Miaojing Shi, Meng Wei et al.

Radiology report generation (RRG) has attracted significant attention due to its potential to reduce the workload of radiologists. Current RRG approaches are still unsatisfactory against clinical standards. This paper introduces a novel RRG method, \textbf{LM-RRG}, that integrates large models (LMs) with clinical quality reinforcement learning to generate accurate and comprehensive chest X-ray radiology reports. Our method first designs a large language model driven feature extractor to analyze and interpret different regions of the chest X-ray image, emphasizing specific regions with medical significance. Next, based on the large model's decoder, we develop a multimodal report generator that leverages multimodal prompts from visual features and textual instruction to produce the radiology report in an auto-regressive way. Finally, to better reflect the clinical significant and insignificant errors that radiologists would normally assign in the report, we introduce a novel clinical quality reinforcement learning strategy. It utilizes the radiology report clinical quality (RadCliQ) metric as a reward function in the learning process. Extensive experiments on the MIMIC-CXR and IU-Xray datasets demonstrate the superiority of our method over the state of the art.

Jianyi Wang, Shanchuan Lin, Zhijie Lin et al.

Recent advances in diffusion-based video restoration (VR) demonstrate significant improvement in visual quality, yet yield a prohibitive computational cost during inference. While several distillation-based approaches have exhibited the potential of one-step image restoration, extending existing approaches to VR remains challenging and underexplored, particularly when dealing with high-resolution video in real-world settings. In this work, we propose a one-step diffusion-based VR model, termed as SeedVR2, which performs adversarial VR training against real data. To handle the challenging high-resolution VR within a single step, we introduce several enhancements to both model architecture and training procedures. Specifically, an adaptive window attention mechanism is proposed, where the window size is dynamically adjusted to fit the output resolutions, avoiding window inconsistency observed under high-resolution VR using window attention with a predefined window size. To stabilize and improve the adversarial post-training towards VR, we further verify the effectiveness of a series of losses, including a proposed feature matching loss without significantly sacrificing training efficiency. Extensive experiments show that SeedVR2 can achieve comparable or even better performance compared with existing VR approaches in a single step.

Meng Wei, Qianyi Wu, Jianmin Zheng et al.

Rendering and reconstruction are long-standing topics in computer vision and graphics. Achieving both high rendering quality and accurate geometry is a challenge. Recent advancements in 3D Gaussian Splatting (3DGS) have enabled high-fidelity novel view synthesis at real-time speeds. However, the noisy and discrete nature of 3D Gaussian primitives hinders accurate surface estimation. Previous attempts to regularize 3D Gaussian normals often degrade rendering quality due to the fundamental disconnect between normal vectors and the rendering pipeline in 3DGS-based methods. Therefore, we introduce Normal-GS, a novel approach that integrates normal vectors into the 3DGS rendering pipeline. The core idea is to model the interaction between normals and incident lighting using the physically-based rendering equation. Our approach re-parameterizes surface colors as the product of normals and a designed Integrated Directional Illumination Vector (IDIV). To optimize memory usage and simplify optimization, we employ an anchor-based 3DGS to implicitly encode locally-shared IDIVs. Additionally, Normal-GS leverages optimized normals and Integrated Directional Encoding (IDE) to accurately model specular effects, enhancing both rendering quality and surface normal precision. Extensive experiments demonstrate that Normal-GS achieves near state-of-the-art visual quality while obtaining accurate surface normals and preserving real-time rendering performance.

Xiaoyu Yue, Zidong Wang, Zeyu Lu et al.

Conventional class-guided diffusion models generally succeed in generating images with correct semantic content, but often struggle with texture details. This limitation stems from the usage of class priors, which only provide coarse and limited conditional information. To address this issue, we propose Diffusion on Diffusion (DoD), an innovative multi-stage generation framework that first extracts visual priors from previously generated samples, then provides rich guidance for the diffusion model leveraging visual priors from the early stages of diffusion sampling. Specifically, we introduce a latent embedding module that employs a compression-reconstruction approach to discard redundant detail information from the conditional samples in each stage, retaining only the semantic information for guidance. We evaluate DoD on the popular ImageNet-$256 \times 256$ dataset, reducing 7$\times$ training cost compared to SiT and DiT with even better performance in terms of the FID-50K score. Our largest model DoD-XL achieves an FID-50K score of 1.83 with only 1 million training steps, which surpasses other state-of-the-art methods without bells and whistles during inference.

Meng Wei, Zhongnian Li, Yong Zhou et al.

Long-tailed data is prevalent in real-world classification tasks and heavily relies on supervised information, which makes the annotation process exceptionally labor-intensive and time-consuming. Unfortunately, despite being a common approach to mitigate labeling costs, existing weakly supervised learning methods struggle to adequately preserve supervised information for tail samples, resulting in a decline in accuracy for the tail classes. To alleviate this problem, we introduce a novel weakly supervised labeling setting called Reduced Label. The proposed labeling setting not only avoids the decline of supervised information for the tail samples, but also decreases the labeling costs associated with long-tailed data. Additionally, we propose an straightforward and highly efficient unbiased framework with strong theoretical guarantees to learn from these Reduced Labels. Extensive experiments conducted on benchmark datasets including ImageNet validate the effectiveness of our approach, surpassing the performance of state-of-the-art weakly supervised methods.

Zhongnian Li, Meng Wei, Peng Ying et al.

Learning from Multi-Positive and Unlabeled (MPU) data has gradually attracted significant attention from practical applications. Unfortunately, the risk of MPU also suffer from the shift of minimum risk, particularly when the models are very flexible as shown in Fig.\ref{moti}. In this paper, to alleviate the shifting of minimum risk problem, we propose an Example Sieve Approach (ESA) to select examples for training a multi-class classifier. Specifically, we sieve out some examples by utilizing the Certain Loss (CL) value of each example in the training stage and analyze the consistency of the proposed risk estimator. Besides, we show that the estimation error of proposed ESA obtains the optimal parametric convergence rate. Extensive experiments on various real-world datasets show the proposed approach outperforms previous methods.

Zhongnian Li, Meng Wei, Peng Ying et al.

Annotating data for sensitive labels (e.g., disease, smoking) poses a potential threats to individual privacy in many real-world scenarios. To cope with this problem, we propose a novel setting to protect privacy of each instance, namely learning from concealed labels for multi-class classification. Concealed labels prevent sensitive labels from appearing in the label set during the label collection stage, which specifies none and some random sampled insensitive labels as concealed labels set to annotate sensitive data. In this paper, an unbiased estimator can be established from concealed data under mild assumptions, and the learned multi-class classifier can not only classify the instance from insensitive labels accurately but also recognize the instance from the sensitive labels. Moreover, we bound the estimation error and show that the multi-class classifier achieves the optimal parametric convergence rate. Experiments demonstrate the significance and effectiveness of the proposed method for concealed labels in synthetic and real-world datasets.

Meng Wei, Zhongnian Li, Peng Ying et al.

Existing similarity-based weakly supervised learning approaches often rely on precise similarity annotations between data pairs, which may inadvertently expose sensitive label information and raise privacy risks. To mitigate this issue, we propose Uncertain Similarity and Unlabeled Learning (USimUL), a novel framework where each similarity pair is embedded with an uncertainty component to reduce label leakage. In this paper, we propose an unbiased risk estimator that learns from uncertain similarity and unlabeled data. Additionally, we theoretically prove that the estimator achieves statistically optimal parametric convergence rates. Extensive experiments on both benchmark and real-world datasets show that our method achieves superior classification performance compared to conventional similarity-based approaches.

Meng Wei, Charlie Budd, Oluwatosin Alabi et al.

Consistent surgical instrument segmentation is critical for automation in robot-assisted surgery. Yet, existing methods only treat instrument-level instance segmentation (IIS) or part-level semantic segmentation (PSS) separately, without interaction between these tasks. In this work, we formulate a surgical tool segmentation as a unified part-aware instance segmentation (PIS) problem and introduce SurgPIS, the first PIS model for surgical instruments. Our method adopts a transformer-based mask classification approach and introduces part-specific queries derived from instrument-level object queries, explicitly linking parts to their parent instrument instances. In order to address the lack of large-scale datasets with both instance- and part-level labels, we propose a weakly-supervised learning strategy for SurgPIS to learn from disjoint datasets labelled for either IIS or PSS purposes. During training, we aggregate our PIS predictions into IIS or PSS masks, thereby allowing us to compute a loss against partially labelled datasets. A student-teacher approach is developed to maintain prediction consistency for missing PIS information in the partially labelled data, e.g., parts of the IIS labelled data. Extensive experiments across multiple datasets validate the effectiveness of SurgPIS, achieving state-of-the-art performance in PIS as well as IIS, PSS, and instrument-level semantic segmentation.

Tobias Rueckert, David Rauber, Raphaela Maerkl et al.

Reliable recognition and localization of surgical instruments in endoscopic video recordings are foundational for a wide range of applications in computer- and robot-assisted minimally invasive surgery (RAMIS), including surgical training, skill assessment, and autonomous assistance. However, robust performance under real-world conditions remains a significant challenge. Incorporating surgical context - such as the current procedural phase - has emerged as a promising strategy to improve robustness and interpretability. To address these challenges, we organized the Surgical Procedure Phase, Keypoint, and Instrument Recognition (PhaKIR) sub-challenge as part of the Endoscopic Vision (EndoVis) challenge at MICCAI 2024. We introduced a novel, multi-center dataset comprising thirteen full-length laparoscopic cholecystectomy videos collected from three distinct medical institutions, with unified annotations for three interrelated tasks: surgical phase recognition, instrument keypoint estimation, and instrument instance segmentation. Unlike existing datasets, ours enables joint investigation of instrument localization and procedural context within the same data while supporting the integration of temporal information across entire procedures. We report results and findings in accordance with the BIAS guidelines for biomedical image analysis challenges. The PhaKIR sub-challenge advances the field by providing a unique benchmark for developing temporally aware, context-driven methods in RAMIS and offers a high-quality resource to support future research in surgical scene understanding.

Meng Wei, Zhongnian Li, Peng Ying et al.

Recent advances in vision-language models (VLMs) have demonstrated remarkable capabilities in image classification by leveraging predefined sets of labels to construct text prompts for zero-shot reasoning. However, these approaches face significant limitations in undefined domains, where the label space is vocabulary-unknown and composite. We thus introduce Generative Semantic Labels (GSLs), a novel task that aims to predict a comprehensive set of semantic labels for an image without being constrained by a predefined labels set. Unlike traditional zero-shot classification, GSLs generates multiple semantic-level labels, encompassing objects, scenes, attributes, and relationships, thereby providing a richer and more accurate representation of image content. In this paper, we propose Chain-of-Action (CoA), an innovative method designed to tackle the GSLs task. CoA is motivated by the observation that enriched contextual information significantly improves generative performance during inference. Specifically, CoA decomposes the GSLs task into a sequence of detailed actions. Each action extracts and merges key information from the previous step, passing enriched context to the next, ultimately guiding the VLM to generate comprehensive and accurate semantic labels. We evaluate the effectiveness of CoA through extensive experiments on widely-used benchmark datasets. The results demonstrate significant improvements across key performance metrics, validating the capability of CoA to generate accurate and contextually rich semantic labels. Our work not only advances the state-of-the-art in generative semantic labels but also opens new avenues for applying VLMs in open-ended and dynamic real-world scenarios.

Meng Wei, Zhongnian Li, Peng Ying et al.

In multi-label classification, each training instance is associated with multiple class labels simultaneously. Unfortunately, collecting the fully precise class labels for each training instance is time- and labor-consuming for real-world applications. To alleviate this problem, a novel labeling setting termed \textit{Determined Multi-Label Learning} (DMLL) is proposed, aiming to effectively alleviate the labeling cost inherent in multi-label tasks. In this novel labeling setting, each training instance is associated with a \textit{determined label} (either "Yes" or "No"), which indicates whether the training instance contains the provided class label. The provided class label is randomly and uniformly selected from the whole candidate labels set. Besides, each training instance only need to be determined once, which significantly reduce the annotation cost of the labeling task for multi-label datasets. In this paper, we theoretically derive an risk-consistent estimator to learn a multi-label classifier from these determined-labeled training data. Additionally, we introduce a similarity-based prompt learning method for the first time, which minimizes the risk-consistent loss of large-scale pre-trained models to learn a supplemental prompt with richer semantic information. Extensive experimental validation underscores the efficacy of our approach, demonstrating superior performance compared to existing state-of-the-art methods.

Meng Wei, Chun Yuan, Xiaoyu Yue et al.

Scene graph generation aims to produce structured representations for images, which requires to understand the relations between objects. Due to the continuous nature of deep neural networks, the prediction of scene graphs is divided into object detection and relation classification. However, the independent relation classes cannot separate the visual features well. Although some methods organize the visual features into graph structures and use message passing to learn contextual information, they still suffer from drastic intra-class variations and unbalanced data distributions. One important factor is that they learn an unstructured output space that ignores the inherent structures of scene graphs. Accordingly, in this paper, we propose a Higher Order Structure Embedded Network (HOSE-Net) to mitigate this issue. First, we propose a novel structure-aware embedding-to-classifier(SEC) module to incorporate both local and global structural information of relationships into the output space. Specifically, a set of context embeddings are learned via local graph based message passing and then mapped to a global structure based classification space. Second, since learning too many context-specific classification subspaces can suffer from data sparsity issues, we propose a hierarchical semantic aggregation(HSA) module to reduces the number of subspaces by introducing higher order structural information. HSA is also a fast and flexible tool to automatically search a semantic object hierarchy based on relational knowledge graphs. Extensive experiments show that the proposed HOSE-Net achieves the state-of-the-art performance on two popular benchmarks of Visual Genome and VRD.

Zhenghang Zhong, Zhe Tang, Xiangxing Li et al.

In this paper, we present a new location fingerprinting database comprised of Wi-Fi received signal strength (RSS) and geomagnetic field intensity measured with multiple devices at a multi-floor building in Xi'an Jiatong-Liverpool University, Suzhou, China. We also provide preliminary results of localization and trajectory estimation based on convolutional neural network (CNN) and long short-term memory (LSTM) network with this database. For localization, we map RSS data for a reference point to an image-like, two-dimensional array and then apply CNN which is popular in image and video analysis and recognition. For trajectory estimation, we use a modified random way point model to efficiently generate continuous step traces imitating human walking and train a stacked two-layer LSTM network with the generated data to remember the changing pattern of geomagnetic field intensity against (x,y) coordinates. Experimental results demonstrate the usefulness of our new database and the feasibility of the CNN and LSTM-based localization and trajectory estimation with the database.