IEEE Transactions on Circuits and Systems for Video Technology最新文献

Stage-Wise Affinity Evolution Spatial Propagation Network for Depth Completion 深度完井阶段亲和演化空间传播网络

IF 10.8 1区工程技术

IEEE Transactions on Circuits and Systems for Video Technology Pub Date : 2026-10-01 Epub Date: 2026-06-19 DOI: 10.1109/TCSVT.2026.3705542

Bin Fan;Zhengyu Zhu;Hongmin Liu

{"title":"Stage-Wise Affinity Evolution Spatial Propagation Network for Depth Completion","authors":"Bin Fan;Zhengyu Zhu;Hongmin Liu","doi":"10.1109/TCSVT.2026.3705542","DOIUrl":"https://doi.org/10.1109/TCSVT.2026.3705542","url":null,"abstract":"Spatial Propagation Networks (SPNs) have demonstrated remarkable effectiveness in RGB-guided depth completion through iterative refinement of depth maps. However, conventional SPN methods, such as CSPN, use fixed affinity maps during propagation, which results in suboptimal performance as static affinities fail to adapt to dynamically evolving depth structures. To address this issue, we propose a novel depth propagation strategy, Affinity Evolution, which dynamically adjusts affinity maps based on the evolving depth map during the propagation process. Unlike previous SPN methods that rely on static, precomputed affinity maps, our approach computes affinity maps directly from the depth map using the Perona–Malik equation, preserving spatial structures and enhancing adaptability. To further improve the adaptability of Affinity Evolution, we design the Depth Map Re-Encoding and Spatial Self-Attention modules to extract and enhance depth-aware feature representations. The resulting model, Stage-wise Affinity Evolution Spatial Propagation Network (SAE-SPN), achieves highly competitive performance on several depth completion datasets. Additionally, extensive experiments under challenging settings, such as extreme sparsity and unguided depth completion, demonstrate the robustness and effectiveness of the proposed method.","PeriodicalId":13082,"journal":{"name":"IEEE Transactions on Circuits and Systems for Video Technology","volume":"36 10","pages":"15023-15038"},"PeriodicalIF":10.8,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149097008","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

TRAN: Texture Response Adjustment Network for Lightweight Single-Image Super-Resolution 轻量级单图像超分辨率纹理响应调整网络

IF 10.8 1区工程技术

IEEE Transactions on Circuits and Systems for Video Technology Pub Date : 2026-10-01 Epub Date: 2026-06-05 DOI: 10.1109/TCSVT.2026.3700195

Ping Zheng;Rongfeng Zhou;He Jiang;Wen Cui;Liangliang Chen;Qiqi Kou;Deqiang Cheng;Guangtao Zhai

{"title":"TRAN: Texture Response Adjustment Network for Lightweight Single-Image Super-Resolution","authors":"Ping Zheng;Rongfeng Zhou;He Jiang;Wen Cui;Liangliang Chen;Qiqi Kou;Deqiang Cheng;Guangtao Zhai","doi":"10.1109/TCSVT.2026.3700195","DOIUrl":"https://doi.org/10.1109/TCSVT.2026.3700195","url":null,"abstract":"Single Image Super-Resolution (SISR) aims to reconstruct a high-resolution (HR) image from a given low-resolution (LR) input. Reconstruction difficulty varies across textures of different complexity. However, existing methods apply a single, unified feature modeling and reconstruction pipeline for all textures. This often exacerbates the reconstruction imbalance across textures of varying complexity. To address this issue, this study proposes Texture Response Adjustment Network for SISR (TRAN). This method achieves more balanced fine-grained reconstruction and improves overall reconstruction quality by adaptively adjusting texture response. TRAN comprises two main components. The first mitigates local texture response bias within an image by decoupling textures of different complexity in the frequency domain for independent modeling and response adjustment. The second alleviates global texture response bias across different images by dynamically perceiving overall image complexity and adjusting the network structure accordingly. Experiments on multiple benchmark datasets demonstrate that the proposed TRAN balances fine-grained reconstruction across textures of different complexity and improve overall reconstruction performance. Notably, on the <inline-formula> <tex-math>$times 4$ </tex-math></inline-formula> SISR task of the Manga109 dataset, TRAN outperforms MambaIRv2-L by 0.31 dB in PSNR while reducing FLOPs by 38.1G.","PeriodicalId":13082,"journal":{"name":"IEEE Transactions on Circuits and Systems for Video Technology","volume":"36 10","pages":"15347-15366"},"PeriodicalIF":10.8,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149097076","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Expanding Intra-Modal and Compressing Inter-Modal Embedding Space for Incomplete Multimodal Emotion Recognition 扩展模态内和压缩模态间嵌入空间的不完全多模态情感识别

IF 10.8 1区工程技术

IEEE Transactions on Circuits and Systems for Video Technology Pub Date : 2026-10-01 Epub Date: 2026-06-12 DOI: 10.1109/TCSVT.2026.3702958

Zhibin Wu;Wenhao Li;Zheyun Qin;Qiangchang Wang

{"title":"Expanding Intra-Modal and Compressing Inter-Modal Embedding Space for Incomplete Multimodal Emotion Recognition","authors":"Zhibin Wu;Wenhao Li;Zheyun Qin;Qiangchang Wang","doi":"10.1109/TCSVT.2026.3702958","DOIUrl":"https://doi.org/10.1109/TCSVT.2026.3702958","url":null,"abstract":"Multimodal Emotion Recognition (MER) often suffers from incomplete modality inputs caused by sensor failures or privacy constraints. Recent studies generally fall into two categories: reconstruction-based methods and joint representation learning. The former recovers missing modalities from available ones, while the latter projects modalities into a shared latent space. However, these methods remain sensitive to single-modality variations and struggle to achieve effective alignment due to the gap between homogeneous intra-modal and heterogeneous inter-modal embedding spaces. To address these limitations, we propose a novel ECI framework that simultaneously Expands intra-modal and Compresses Inter-modal embedding spaces. ECI consists of three key components: Inter-modal Volume Alignment (IVA), HyperGraph-augmented Self-attention (HGS), and Intra-modal Distribution Normalization (IDN). Firstly, IVA reduces geometric inconsistency by minimizing the kernelized volume they form, encouraging compact and well-aligned cross-modal distributions. Secondly, HGS integrates hypergraph convolution with self-attention to capture high-order dependencies and pairwise interactions, enabling effective reconstruction of missing modalities. Finally, IDN applies contrastive suppression to expand feature distributions within intra-modal spaces to avoid overfitting and strengthen the discriminative power of each modality. Consequently, ECI achieves global alignment and complementary information flow while enhancing individual modality discriminability. Experiments on multiple benchmark datasets demonstrate that ECI achieves state-of-the-art performance across various missing modality scenarios.","PeriodicalId":13082,"journal":{"name":"IEEE Transactions on Circuits and Systems for Video Technology","volume":"36 10","pages":"15568-15580"},"PeriodicalIF":10.8,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149097091","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

D2-CDIG: Controlled Diffusion Remote Sensing Image Generation With Dual Priors of DEM and Cloud-Fog 2d - cdig：基于DEM和云雾双先验的可控扩散遥感图像生成

IF 10.8 1区工程技术

IEEE Transactions on Circuits and Systems for Video Technology Pub Date : 2026-10-01 Epub Date: 2026-07-03 DOI: 10.1109/TCSVT.2026.3709817

Zuopeng Zhao;Ying Liu;Kanyaphakphachsorn Pharksuwan;Su Luo;Xiaoyu Li;Maocai Ning

{"title":"D2-CDIG: Controlled Diffusion Remote Sensing Image Generation With Dual Priors of DEM and Cloud-Fog","authors":"Zuopeng Zhao;Ying Liu;Kanyaphakphachsorn Pharksuwan;Su Luo;Xiaoyu Li;Maocai Ning","doi":"10.1109/TCSVT.2026.3709817","DOIUrl":"https://doi.org/10.1109/TCSVT.2026.3709817","url":null,"abstract":"Remote sensing image generation provides a reliable data foundation for remote sensing large models and downstream tasks. However, existing controllable remote sensing image generation methods typically rely on traditional techniques such as segmentation and edge detection, which do not fully leverage terrain or atmospheric conditions. As a result, the generated images often lack accuracy and naturalness when dealing with complex terrains and atmospheric phenomena. In this paper, we propose a novel remote sensing image generation framework, D2-CDIG, which integrates diffusion models with a dual-prior control mechanism. By incorporating both Digital Elevation Model (DEM) and cloud-fog information as dual prior knowledge, D2-CDIG precisely controls ground features and atmospheric phenomena within the generated images. Specifically, D2-CDIG decouples the terrain and atmospheric generation processes through independent control of ground and atmospheric branches. Additionally, a refined cloud-fog slider is introduced to flexibly adjust cloud thickness and distribution. During training, ground and atmospheric control signals are injected in layers to ensure a seamless transition within the images. Compared to traditional methods based on segmentation or edge detection, D2-CDIG shows significant improvements in image quality, detail richness, and realism. D2-CDIG offers a flexible and precise solution for remote sensing image generation, providing high-quality data for training large remote sensing models and downstream tasks.","PeriodicalId":13082,"journal":{"name":"IEEE Transactions on Circuits and Systems for Video Technology","volume":"36 10","pages":"14696-14709"},"PeriodicalIF":10.8,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149097248","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Mesh Denoising Transformer 网格去噪变压器

IF 10.8 1区工程技术

IEEE Transactions on Circuits and Systems for Video Technology Pub Date : 2026-10-01 Epub Date: 2026-07-06 DOI: 10.1109/TCSVT.2026.3710517

Wenbo Zhao;Xianming Liu;Deming Zhai;Junjun Jiang;Xiangyang Ji

{"title":"Mesh Denoising Transformer","authors":"Wenbo Zhao;Xianming Liu;Deming Zhai;Junjun Jiang;Xiangyang Ji","doi":"10.1109/TCSVT.2026.3710517","DOIUrl":"https://doi.org/10.1109/TCSVT.2026.3710517","url":null,"abstract":"Mesh denoising, aimed at removing noise from input meshes while preserving their feature structures, is a practical yet challenging task. Despite the remarkable progress in learning-based mesh denoising methods in recent years, their network designs still suffer from two major limitations: a dependence on single-modal geometric representations, which fall short of capturing the multifaceted attributes of meshes, and a lack of effective global feature aggregation, which limits their ability to model the global mesh structure. To tackle these issues, we propose <italic>SurfaceFormer, a novel Transformer-based mesh denoising framework. Our first contribution is the development of a new representation known as <italic>Local Surface Descriptor, which is crafted by establishing polar coordinate systems on each mesh face, followed by sampling points from adjacent surfaces using geodesics. The normals of the sampled points are arranged into image-like 2D patches to capture local geometric details, whereas the poles and vertex coordinates are consolidated into a point cloud to represent spatial information. This representation alleviates the challenges posed by the irregular and non-Euclidean characteristics of mesh data, facilitating a smooth integration with the Transformer architecture. Next, we propose a dual-stream structure consisting of a <italic>Geometric Encoder branch and a <italic>Spatial Encoder branch, which jointly encode local geometry details and spatial information to fully explore multimodal information for mesh denoising. A subsequent <italic>Denoising Transformer module receives the multimodal information and achieves efficient global feature aggregation through self-attention operators. Our experimental evaluations demonstrate that this novel approach outperforms existing state-of-the-art methods in both objective and subjective assessments, representing a significant advance in mesh denoising.","PeriodicalId":13082,"journal":{"name":"IEEE Transactions on Circuits and Systems for Video Technology","volume":"36 10","pages":"15610-15623"},"PeriodicalIF":10.8,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149097229","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

RAPNet: Rate-Distortion-Aware Video Pre-Processing via One-Step Reinforcement Learning for Video Coding RAPNet：基于一步强化学习的视频编码的率失真感知视频预处理

IF 10.8 1区工程技术

IEEE Transactions on Circuits and Systems for Video Technology Pub Date : 2026-10-01 Epub Date: 2026-06-05 DOI: 10.1109/TCSVT.2026.3700718

Yanchen Zhao;Jiaye Fu;Zhimeng Huang;Jiaqi Zhang;Siwei Ma

{"title":"RAPNet: Rate-Distortion-Aware Video Pre-Processing via One-Step Reinforcement Learning for Video Coding","authors":"Yanchen Zhao;Jiaye Fu;Zhimeng Huang;Jiaqi Zhang;Siwei Ma","doi":"10.1109/TCSVT.2026.3700718","DOIUrl":"https://doi.org/10.1109/TCSVT.2026.3700718","url":null,"abstract":"In recent years, with the rapid development of Neural Network-based (NN-based) deep learning algorithms, NN-based video coding tools have been emerging, demonstrating excellent coding performance. However, constrained by limited computational power at the edge or on end-user devices, deploying high-performance NN-based coding tools is a challenging task. This limitation hinders the standardization and widespread application of NN-based coding tools. Recently, NN-based video pre-processing has gained significant attention from both industry and academia. These algorithms are deployed on the cloud and require no modifications to the video codec. Traditional pre-processing methods typically rely on handcrafted designs or natural degradation modeling for quality enhancement. These approaches cannot be jointly optimized with the encoder, making it difficult to perceive the encoding process or adapt to different video content. Moreover, the non-differentiable nature of real codecs makes conventional supervised optimization difficult. Therefore, we propose a one-step reinforcement learning-based framework for video pre-processing model training. We also develop a video pre-processing network capable of supporting 4K@60fps live streaming applications. Experimental results demonstrate that our method achieves significant performance improvements across various video codecs. Compared with codecs such as <italic>x265, <italic>VVenC, and <italic>SVT-AVS3, our method achieves more than a 20% BD-rate (VMAF) improvement. The inference code and pretrained model will be released at: <uri>https://github.com/pkuyczhao/RAPNet</uri>","PeriodicalId":13082,"journal":{"name":"IEEE Transactions on Circuits and Systems for Video Technology","volume":"36 10","pages":"15818-15831"},"PeriodicalIF":10.8,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149097257","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

The Devil Is in the Masks: Label Disentangling Learning Network for Camouflaged Object Detection 魔鬼在面具里：用于伪装对象检测的标签解缠学习网络

IF 10.8 1区工程技术

IEEE Transactions on Circuits and Systems for Video Technology Pub Date : 2026-10-01 Epub Date: 2026-06-17 DOI: 10.1109/TCSVT.2026.3704621

Xiandong Wang;Fengqin Yao;Qi Guo;Tianqi Guo;Shengke Wang;Guoqiang Zhong;Qing Cai;Junyu Dong

{"title":"The Devil Is in the Masks: Label Disentangling Learning Network for Camouflaged Object Detection","authors":"Xiandong Wang;Fengqin Yao;Qi Guo;Tianqi Guo;Shengke Wang;Guoqiang Zhong;Qing Cai;Junyu Dong","doi":"10.1109/TCSVT.2026.3704621","DOIUrl":"https://doi.org/10.1109/TCSVT.2026.3704621","url":null,"abstract":"Camouflaged object detection involves identifying objects that closely resemble their background. Despite the significant progress made by existing methods, we observe that detecting camouflaged object <italic>boundaries remains more challenging than predicting central regions. This difficulty primarily arises from the use of binary mask supervision during training, which treats all pixels uniformly and neglects the varying importance of different pixels. To address this problem, we first employ distance transformation to disentangle the mask into a body map and a detail map. The body map focuses on the central region of the object, while the detail map concentrates on boundary areas, capturing more comprehensive pixel information than traditional boundary supervision methods. Building on this, we propose a label disentangling learning network, LDLNet, which consists of a body branch for body map learning, a detail branch for detail map capturing, and an iterative interaction branch to enhance collaboration between body and detail features. Specifically, to alleviate redundant information generated during the fusion process at different levels, body branch introduces a multi-dimensional denoising fusion block, whereas detail branch uses a multi-view context awareness block to capture the fragile boundaries of camouflaged objects. The iterative interaction branch fuses body and detail features and feeds them back to refine feature extraction in each branch. Extensive experimental results on four widely used public datasets demonstrate that LDLNet outperforms the state-of-the-art approaches. Code is available at <uri>https://github.com/XDWang97/LDLNet</uri>","PeriodicalId":13082,"journal":{"name":"IEEE Transactions on Circuits and Systems for Video Technology","volume":"36 10","pages":"14740-14754"},"PeriodicalIF":10.8,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149097298","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

HEPE++: Toward VLM-Based Hybrid Explainable Prompt Enhancement for Zero-Shot Industrial Anomaly Detection hepe++：基于vlm的混合可解释提示增强零射击工业异常检测

IF 10.8 1区工程技术

IEEE Transactions on Circuits and Systems for Video Technology Pub Date : 2026-10-01 Epub Date: 2026-06-05 DOI: 10.1109/TCSVT.2026.3700398

Weichao Cai;Weiliang Huang;Yunkang Cao;Chao Huang;Bob Zhang;Fei Yuan;Jie Wen;Weiming Shen

{"title":"HEPE++: Toward VLM-Based Hybrid Explainable Prompt Enhancement for Zero-Shot Industrial Anomaly Detection","authors":"Weichao Cai;Weiliang Huang;Yunkang Cao;Chao Huang;Bob Zhang;Fei Yuan;Jie Wen;Weiming Shen","doi":"10.1109/TCSVT.2026.3700398","DOIUrl":"https://doi.org/10.1109/TCSVT.2026.3700398","url":null,"abstract":"Zero-Shot Industrial Anomaly Detection (ZSIAD) aims to identify and localize anomalies in industrial images from unseen categories. Owing to the powerful generalization capabilities, Vision-Language Models (VLMs) have achieved growing interest in ZSIAD. To guide the model toward understanding and localizing the semantically complex industrial anomalies, existing VLM-based methods have attempted to provide additional prompts to the model through learnable text prompt templates. However, these zero-shot methods lack detailed descriptions of specific anomalies, making it difficult to classify and segment the diverse range of industrial anomalies accurately. To address the aforementioned issue, we propose an effective solution, which includes a multi-stage prompt generation agent and a Hybrid Explainable Prompt Enhancement plus (HEPE++) framework. Specifically, we leverage the Multi-modal Language Large Model (MLLM) to articulate the detailed differential information between normal and test samples, which can provide detailed text prompts to the model through further refinement and anti-false alarm constraint. Moreover, we introduce the Visual Fundamental Model (VFM) to generate anomaly-related attention prompts for more accurate localization of anomalies with varying sizes and shapes. Building on this foundation, HEPE++ jointly utilizes generated detailed text and attention prompts to enhance the performance of VLM-based methods in ZSIAD. Furthermore, we introduce a novel gated attention mechanism to augment the model’s ability to focus on potential subtle anomalous regions in the foreground. Extensive experiments on seven real-world industrial anomaly detection datasets have shown that the proposed method not only outperforms recent SOTA methods, but also its explainable prompts provide the model with a more intuitive basis for anomaly identification.","PeriodicalId":13082,"journal":{"name":"IEEE Transactions on Circuits and Systems for Video Technology","volume":"36 10","pages":"14867-14881"},"PeriodicalIF":10.8,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149097328","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Flow6D: Lie-Group Consistent Riemannian Flow-Matching for Category-Level 6D Object Pose and Size Estimation Flow6D：类别级6D对象姿态和大小估计的liegroup一致riemanian流匹配

IF 10.8 1区工程技术

IEEE Transactions on Circuits and Systems for Video Technology Pub Date : 2026-10-01 Epub Date: 2026-06-24 DOI: 10.1109/TCSVT.2026.3706824

Kai Sun;Jin Liu;Yunfeng Kang;Yanzi Miao;Hesheng Wang

{"title":"Flow6D: Lie-Group Consistent Riemannian Flow-Matching for Category-Level 6D Object Pose and Size Estimation","authors":"Kai Sun;Jin Liu;Yunfeng Kang;Yanzi Miao;Hesheng Wang","doi":"10.1109/TCSVT.2026.3706824","DOIUrl":"https://doi.org/10.1109/TCSVT.2026.3706824","url":null,"abstract":"Category-level object pose and size estimation is a critical capability for robotic manipulation in unstructured 3D environments. However, existing methods struggle to maintain stable and accurate pose predictions when handling intra-class shape variations and complex spatial transformations. To address this challenge, we propose Flow6D, a Riemannian flow matching-based framework for category-level pose and size estimation, trained solely on synthetic data. Specifically, Flow6D models pose transformations on the SE(3) Riemannian manifold, explicitly preserving the compositionality and continuity of rotations and translations, thereby improving generalization to unseen objects. To further enhance geometric equivariance, we introduce a Lie-group consistency mechanism that constructs a closed transformation set in the Sim(3) group and incorporates group structure constraints, enabling the learning of composable, invertible, and continuous geometric transformations. Additionally, Flow6D integrates a geometric prediction and encoding module and a conditional attention mechanism, which respectively refine geometric feature extraction and dynamically adapt feature interactions, improving pose estimation accuracy and robustness. Extensive quantitative and qualitative experiments on REAL275 and Wild6D, as well as real-world robotic grasping tasks with the Franka robotic arm, demonstrate the effectiveness and practical applicability of Flow6D.","PeriodicalId":13082,"journal":{"name":"IEEE Transactions on Circuits and Systems for Video Technology","volume":"36 10","pages":"14416-14430"},"PeriodicalIF":10.8,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149097360","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

SceneVLP: Augmenting Vision-Language Pre-Training Models for Video Action Recognition With Video Scene Graphs 基于视频场景图的视频动作识别增强视觉语言预训练模型

IF 10.8 1区工程技术

IEEE Transactions on Circuits and Systems for Video Technology Pub Date : 2026-10-01 Epub Date: 2026-06-09 DOI: 10.1109/TCSVT.2026.3701841

Liangzhen Liu;Guangzhao Li;Qi Xie;Zhaowen Wang;Li Kuang;Yingjie Xia;Jianwei Yin

{"title":"SceneVLP: Augmenting Vision-Language Pre-Training Models for Video Action Recognition With Video Scene Graphs","authors":"Liangzhen Liu;Guangzhao Li;Qi Xie;Zhaowen Wang;Li Kuang;Yingjie Xia;Jianwei Yin","doi":"10.1109/TCSVT.2026.3701841","DOIUrl":"https://doi.org/10.1109/TCSVT.2026.3701841","url":null,"abstract":"Vision-language pre-training (VLP) models have demonstrated remarkable performance in video action recognition. However, their reliance on global video-text alignment often overlooks fine-grained scene details, leading to mis-recognition in visually similar contexts. To address this, we first explore the utilization of fine-grained scene information from video scene graphs (VSGs) to augment video action recognition. First, we introduce FVSG-LM, a frame-aware VSG generation method, which integrates the large model capabilities including MLLM and LLM and employs an inter-frame entity matching algorithm for unsupervised VSG generation. Second, we propose SceneVLP, an augmentation framework which improves VLP models by capturing key events from VSGs with the incorporation of a video key event generation component VKEG, to improve action recognition accuracy. We generate reusable VSGs for mainstream video action recognition datasets, including Kinetics-400, Something-Something-v2, HMDB51, and UCF101. Experimental results demonstrate that the SceneVLP framework, augmented with VSGs, achieves outstanding video action recognition performance, surpassing state-of-the-art methods of similar scale.","PeriodicalId":13082,"journal":{"name":"IEEE Transactions on Circuits and Systems for Video Technology","volume":"36 10","pages":"14491-14503"},"PeriodicalIF":10.8,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149097366","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0