Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileAcademic lineage
Possible advisorsa guess from early papers, not confirmed
- Long QuanSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works465 from public data
- Deep High-Resolution Representation Learning for Human Pose Estimation5,644
This paper proposes a network that maintains high-resolution representations through the whole process of human pose estimation and empirically demonstrates the effectiveness of the network through the superior pose estimation results over two benchmark datasets: the COCO keypoint detection dataset and the MPII Human Pose dataset.
- Deep High-Resolution Representation Learning for Visual Recognition5,145
The superiority of the proposed HRNet in a wide range of applications, including human pose estimation, semantic segmentation, and object detection, is shown, suggesting that the HRNet is a stronger backbone for computer vision problems.
- 4,660
- 1,777
- Object-Contextual Representations for Semantic Segmentation1,756
This paper addresses the semantic segmentation problem with a focus on the context aggregation strategy, and presents a simple yet effective approach, object-contextual representations, characterizing a pixel by exploiting the representation of the corresponding object class.
- Semi-Supervised Semantic Segmentation with Cross Pseudo Supervision1,250
This paper proposes a novel consistency regularization approach, called cross pseudo supervision (CPS), which imposes the consistency on two segmentation networks perturbed with different initialization for the same input image.
- 1,147
- Conditional DETR for Fast Training Convergence1,117
The approach, named conditional DETR, learns a conditional spatial query from the decoder embedding for decoder multi-head cross-attention, which narrows down the spatial range for localizing the distinct regions for object classification and box regression, thus relaxing the dependence on the content embeddings and easing the training.
- A Survey on Learning to Hash1,099
This paper presents a comprehensive survey of the learning to hash algorithms, categorize them according to the manners of preserving the similarities into: pairwise similarity preserving, multiwise Similarity preserving, implicit similarity preserve, as well as quantization, and discusses their relations.
- MARS: A Video Benchmark for Large-Scale Person Re-Identification1,088
It is shown that CNN in classification mode can be trained from scratch using the consecutive bounding boxes of each identity, and the learned CNN embedding outperforms other competing methods considerably and has good generalization ability on other video re-id datasets upon fine-tuning.
- HigherHRNet: Scale-Aware Representation Learning for Bottom-Up Human Pose Estimation952
HigherHRNet is presented, a novel bottom-up human pose estimation method for learning scale-aware representations using high-resolution feature pyramids that surpasses all top-down methods on CrowdPose test and achieves new state-of-the-art result on COCO test-dev, suggesting its robustness in crowded scene.
- Deeply-Learned Part-Aligned Representations for Person Re-identification813
This paper proposes a simple yet effective human part-aligned representation for handling the body part misalignment problem, and shows state-of-the-art results over standard datasets, Market-1501,CUHK03, CUHK01 and VIPeR.
- Structured Knowledge Distillation for Semantic Segmentation748
This paper starts from the straightforward scheme, pixel-wise distillation, which applies the distillation scheme originally introduced for image classification and performs knowledge distillation for each pixel separately, and proposes to distill the structured knowledge from cumbersome networks into compact networks, motivated by the fact that semantic segmentation is a structured prediction problem.
- 668
- 615
- 595
- Part-Aligned Bilinear Representations for Person Re-identification557
A novel network that learns a part-aligned representation for person re-identification that handles the body part misalignment problem, that is, body parts are misaligned across human detections due to pose/viewpoint change and unreliable detection.
- DeepSaliency: Multi-Task Deep Neural Network Model for Salient Object Detection538
This paper proposes a multi-task deep saliency model based on a fully convolutional neural network with global input (whole raw images) and global output (Whole saliency maps) and presents a graph Laplacian regularized nonlinear regression model for saliency refinement.
- Context Autoencoder for Self-supervised Representation Learning522
The CAE design encourages the separation of learning the encoder (representation) from completing the pertaining tasks: masked representation prediction and masked patch reconstruction tasks, and making predictions in the encoded representation space empirically shows the benefit to representation learning.
- Lite-HRNet: A Lightweight High-Resolution Network506
An efficient high-resolution network, Lite-HRNet, is presented, which demonstrates superior results on human pose estimation over popular lightweight networks and can be easily applied to semantic segmentation task in the same lightweight manner.
- Automatic salient object segmentation based on context and shape prior497
A novel automatic salient object segmentation algorithm which integrates both bottom-up salient stimuli and object-level shape prior, leading to binary segmentation of the salient object.
- 496
- Bottom-Up Human Pose Estimation Via Disentangled Keypoint Regression398
It is empirically show that the proposed direct regression method outperforms keypoint detection and grouping methods and achieves superior bottom-up pose estimation results on two benchmark datasets, COCO and CrowdPose.
- Generation of cloned transgenic pigs rich in omega-3 fatty acids386
The generation of cloned pigs that express a humanized Caenorhabditis elegans gene, fat-1, encoding an n-3 fatty acid desaturase is described, and the hfat-1 transgenic pigs produce high levels of n- 3 fatty acids from n-6 analogs, and their tissues have a significantly reduced ratio ofn-6/n-4 fatty acids.
- 381
- Closed-Loop Matters: Dual Regression Networks for Single Image Super-Resolution352
A dual regression scheme is proposed by introducing an additional constraint on LR data to reduce the space of the possible functions and it is shown that this scheme can easily adapt SR models to real-world data, e.g., raw video frames from YouTube.
- Detection of Co-salient Objects by Looking Deep and Wide328
A unified co-salient object detection framework is proposed by introducing two novel insights: looking deep to transfer higher-level representations by using the convolutional neural network with additional adaptive layers could better reflect the sematic properties of the co- salient objects.
- Image-based plant modeling327
A semi-automatic technique for modeling plants directly from images, automating the process of shape recovery while relying on the user to provide simple hints on segmentation, which inherits the realistic shape and complexity of a real plant.
- Admix: Enhancing the Transferability of Adversarial Attacks310
A new input transformation based attack method called Admix that considers the input image and a set of images randomly sampled from other categories that could achieve significantly better transferability than existing input transformation methods under both single model setting and ensemble-model setting.
- 308
- Group DETR: Fast DETR Training with Group-Wise One-to-Many Assignment283
Group DETR is introduced, a simple yet efficient DETR training approach that introduces a group-wise way for one-to-many assignment that speeds up the training convergence and improves the performance of various DETR-based models.
- Forgery-aware Adaptive Transformer for Generalizable Synthetic Image Detection282
This paper develops a forgery-aware adapter to adapt image features to discern and inte-grate local forgery traces within image and frequency do-mains, and finds that considering the contrastive ob-jectives between adapted image features and text prompt embeddings results in a nontrivial generalization improvement.
- Image-based tree modeling278
This paper proposes an approach for generating 3D models of natural-looking trees from images that has the additional benefit of requiring little user intervention and uses the shape patterns of visible branches to predict those of obscured branches.
- DisturbLabel: Regularizing CNN on the Loss Layer266
An extremely simple algorithm which randomly replaces a part of labels as incorrect values in each iteration, which prevents the network training from over-fitting by implicitly averaging over exponentially many networks which are trained with different label sets.
- OCNet: Object Context for Semantic Segmentation262
An efficient interlaced sparse self-attention scheme to model the dense relations between any two of all pixels via the combination of two sparse relation matrices is proposed and empirically shows the advantages of this approach with competitive performances on five challenging benchmarks.
- Composite Quantization for Approximate Nearest Neighbor Search259
This paper presents a novel compact coding approach, composite quantization, for approximate nearest neighbor search, to use the composition of several elements selected from the dictionaries to accurately approximate a vector and to represent the vector by a short code composed of the indices of the selected elements.
- SegFix: Model-Agnostic Boundary Refinement for Segmentation257
A model-agnostic post-processing scheme to improve the boundary quality for the segmentation result that is generated by any existing segmentation model and empirically verify that the SegFix consistently reduces the boundary errors for segmentation results generated from various state-of-the-art models.
- 256
- Cross View Fusion for 3D Human Pose Estimation255
This work introduces a cross-view fusion scheme into CNN to jointly estimate 2D poses for multiple views and presents a recursive Pictorial Structure Model to recover the 3D pose from the multi-view 2D pose.
- Clinical characteristics of 145 patients with corona virus disease 2019 (COVID-19) in Taizhou, Zhejiang, China253
Older patients or patients with comorbidities such as obesity or diabetes mellitus were more likely to have severe condition and more clinical trials are needed.
- An interactive approach to semantic modeling of indoor scenes with an RGBD camera249
An interactive approach to semantic modeling of indoor scenes with a consumer-level RGBD camera, which takes an RGBD image of an indoor scene, which is automatically segmented into a set of regions with semantic labels.
- 248
- Rethinking the Open-Loop Evaluation of End-to-End Autonomous Driving in nuScenes222
An MLP-based method that takes raw sensor data as input and directly outputs the future trajectory of the ego vehicle, without using any perception or prediction information such as camera images or LiDAR, achieves similar end-to-end planning performance on the nuScenes dataset with other perception-based methods.
- 222
- 220
- 220
- Weakly-Shared Deep Transfer Networks for Heterogeneous-Domain Knowledge Propagation214
This paper develops a novel deep network structure, capable of transferring labeling information across heterogeneous domains, especially from text domain to image domain, and presents a novel architecture of DTNs to translate cross-domain information from text to image.
- A simplified method for analysis of polyunsaturated fatty acids213
The lipid extraction procedure prior to methylation employed conventionally in lipid analysis can be omitted without affecting the recovery of long chain (≥ 18 C) fatty acids and their composition.
- 196
- 186
- Weakly-Supervised Action Localization by Generative Attention Modeling184
This paper proposes to model the class-agnostic frame-wise probability conditioned on the frame attention using conditional Variational Auto-Encoder (VAE), and demonstrates advantage of the method and effectiveness in handling action-context confusion problem.
- A Probabilistic Approach to Robust Matrix Factorization178
Experiments show that PRMF is comparable to other state-of-the-art robust matrix factorization methods in terms of accuracy and outperforms them particularly for large data matrices.
- Delicate Textured Mesh Recovery from NeRF via Adaptive Surface Refinement177
A novel framework that generates textured surface meshes from images by efficiently initializing the geometry and view-dependency decomposed appearance with a NeRF and adapting both vertex positions and face density based on reprojected rendering errors is presented.
- 176
- Person Re-Identification with Correspondence Structure Learning175
A boosting-based approach to learn a correspondence structure which indicates the patch-wise matching probabilities between images from a target camera pair is introduced, which integrates a global matching constraint over the learned correspondence structure to exclude cross-view misalignments during the image patch matching process, hence achieving a more reliable matching score between images.
- Cap4Video: What Can Auxiliary Captions Do for Text-Video Retrieval?171
Cap4Video is a novel approach to text-video retrieval, where associated captions from videos are generated using zero-shot video captioning with knowledge from web-scale pre-trained models to complement the original Query-Video matching branch for text- video retrieval.
- 168
- 165
- Quantized correlation hashing for fast cross-modal search165
This work presents a cross-modal hashing approach, called quantized correlation hashing (QCH), which takes into consideration the quantization loss over domains and the relation between domains, and outperforms the state-of-the-art multi- modal hashing methods.
- 165
- Graph-based semi-supervised learning with multiple labels164
A novel graph-based learning framework in the setting of semi-supervised learning with multiple labels is proposed, characterized by simultaneously exploiting the inherent correlations among multiple labels and the label consistency over the graph.
- Bayesian video search reranking161
Two new methods are proposed in this paper to measure the ranking distance based on the disagreement in terms of pair-wise orders, which represents the disagreement between the objective ranking list and the initial text-based.
- Point-Set Anchors for Object Detection, Instance Segmentation and Pose Estimation158
This work argues that the image features extracted at a central point contain limited information for predicting distant keypoints or bounding box boundaries, due to object deformation and scale/orientation variation, and proposes to perform regression from a set of points placed at more advantageous positions to facilitate inference.
- Augmentation Matters: A Simple-Yet-Effective Approach to Semi-Supervised Semantic Segmentation156
This work proposes AugSeg, a simple and clean approach that focuses mainly on data perturbations to boost the SSS performance, and adopts a simplified intensity-based augmentation that selects a random number of data transformations with uniformly sampling distortion strengths from a continuous space.
- Efficient Semantic Video Segmentation with Per-Frame Inference154
This work process efficient semantic video segmentation in a per-frame fashion during the inference process, and explicitly considers the temporal consistency among frames as extra constraints during the training process and embed theporal consistency into the segmentation network.
- MixFormer: Mixing Features across Windows and Dimensions152
This work combines local-window self-attention with depth-wise convolution in a parallel design, modeling cross-window connections to enlarge the receptive fields and proposes bi-directional interactions across branches to provide complementary clues in the channel and spatial dimensions.
- 152
- 149
- Implicit Sample Extension for Unsupervised Person Re-Identification144
An Implicit Sample Extension method to generate support samples from actual samples and their neighbouring clusters in the embedding space through a progressive linear interpolation (PLI) strategy, which reduces the “sub and mixed” clustering errors, thus improving the Re-ID performance.
- Generalized Deep Transfer Networks for Knowledge Propagation in Heterogeneous Domains144
A novel generalized deep transfer networks (DTNs) capable of transferring label information across heterogeneous domains, textual domain to visual domain, and to share the labels between two domains are proposed, able to generate domain-specific and shared interdomain features.
- A Survey of Reasoning with Foundation Models: Concepts, Methodologies, and Outlook142
This article introduces seminal foundation models proposed or adaptable for reasoning, highlighting the latest advancements in various reasoning tasks, methods, and benchmarks, and delves into the potential future directions behind the emergence of reasoning abilities within foundation models.
- Interleaved Structured Sparse Convolutional Neural Networks130
This paper presents a modularized building block, IGC-V2: interleaved structured sparse convolutions, which generalizes interleaves group convolutions to the product of more structured sparse kernels, further eliminating the redundancy.
- StyleSync: High-Fidelity Generalized and Personalized Lip Sync in Style-Based Generator125
This paper designs a mask-guided spatial information encoding module that preserves the details of the given face and enables personalized lip-sync by introducing style space and generator refinement on only limited frames, so the identity and talking style of a target person could be accurately preserved.
- 124
- Optimized Cartesian K-Means122
This paper proposes a novel approach, named optimized cartesian K-means (ock-mean), to better encode the data points for more accurate approximate nearest neighbor search, which can provide more flexibility and lower distortion errors than traditional methods.
- Bidirectional Cross-Modal Knowledge Exploration for Video Recognition with Pre-trained Vision-Language Models121
A novel framework called BIKE, which utilizes the cross-modal bridge to explore bidirectional knowledge and presents a Temporal Concept Spotting mechanism that uses the Text-to-Video expertise to capture temporal saliency in a parameter-free manner, leading to enhanced video representation.
- 121
- Order preserving hashing for approximate nearest neighbor search120
A novel method to learn similarity-preserving hash functions for approximate nearest neighbor (NN) search by maximizing the alignment between the similarity orders computed from the original space and the ones in the hamming space.
- Boosting Adversarial Transferability through Enhanced Momentum119
An enhanced momentum iterative gradient-based method to further enhance the adversarial transferability by accumulating the average gradient of the data points sampled in the gradient direction of the previous iteration so as to stabilize the update direction and escape from poor local maxima.
- 118
- Object Detection in Videos by High Quality Object Linking113
This paper proposes a short tubelet linking algorithm that links temporally-overlapping short tubelets to form long tubelets and improves results by 8.8 percent over the static image detector for fast moving objects.
- Clickage111
It is argued that the massive amount of click data from commercial search engines provides a data set that is unique in the bridging of the semantic and intent gap, and preliminary studies on the power of large-scale click data are presented.
- Fast Approximate $$K$$-Means via Cluster Closures109
Using complexity analysis, real data clustering, and applications to image retrieval, this approach out-performs state-of-the-art approximate k-means algorithms in terms of clustering quality and efficiency.
- Deep Triplet Quantization107
Deep Triplet Quantization (DTQ), a novel approach to learning deep quantization models from the similarity triplets, can generate high-quality and compact binary codes, which yields state-of-the-art image retrieval performance on three benchmark datasets, NUS-WIDE, CIFAR-10, and MS-COCO.
- Results of the NeurIPS'21 Challenge on Billion-Scale Approximate Nearest Neighbor Search104
The outcome of the competition was ranked leaderboards of algorithms in each track based on recall at a query throughput threshold, which compares ANNS algorithms at billion-scale by hardware cost, accuracy and performance.
- UATVR: Uncertainty-Adaptive Text-Video Retrieval102
This paper proposes an Uncertainty-Adaptive Text-Video Retrieval approach, termed UATVR, which models each lookup as a distribution matching procedure and adds additional learnable tokens in the encoders to adaptively aggregate multi-grained semantics for flexible high-level reasoning.
- 99
- MS-DETR: Efficient DETR Training with Mixed Supervision97
Experimental results show that the MS-DETR approach outperforms related DETR variants, such as DN-DETR, Hybrid DETR, and Group DETR, and the combination with related DETR variants further improves the performance.
- LW-DETR: A Transformer Replacement to YOLO for Real-Time Detection97
This paper improves the ViT encoder by aggregating multi-level feature maps, and the intermediate and final feature maps in the ViT encoder, forming richer feature maps, and introduces window-major feature map organization for improving the efficiency of interleaved attention computation.
- S4Net: Single stage salient-instance segmentation97
This paper designs a single stage salient instance segmentation framework, with a novel segmentation branch that regards not only local context inside each detection window but also the surrounding context, enabling us to distinguish instances in the same scope even with partial occlusion.
- 96
- Tag refinement by regularized LDA96
A novel graphical model, regularized Latent Dirichlet Allocation (rLDA), is presented, which facilitates the topic modeling by exploiting both the statistics of tags and visual affinities of images in the corpus.
- 95
- 94
- Accurate Image Search with Multi-Scale Contextual Evidences92
This paper proposes to integrate discriminative cues from multiple contextual levels, i.e., local, regional, and global, via probabilistic analysis, and shows that CNN feature is complementary to SIFT due to its semantic awareness and compares favorably to several other descriptors such as GIST, HSV, etc.
- 91
- Forward Flow for Novel View Synthesis of Dynamic Scenes90
Thorough experiments show that the proposed neural radiance field approach outperforms existing methods in both novel view rendering and motion modeling, demonstrating the effectiveness of the forward flow motion modeling.
- Few-Shot Font Generation by Learning Fine-Grained Local Styles89
A new font generation approach by learning the fine-grained local styles from references, and the spatial correspondence between the content and reference glyphs, which outperforms the state-of-the-art methods in FFG.
- Semi-DETR: Semi-Supervised Object Detection with Detection Transformers87
This work proposes a Stage-wise Hybrid Matching strategy that combines the one-to-many assignment and one- to-one assignment strategies to improve the training efficiency of the first stage and thus provide high-quality pseudo labels for the training of the second stage.
- Human-Object Interaction Detection via Disentangled Transformer87
This work decouple the triplet prediction into human-object pair detection and interaction classification, and presents Disentangled Transformer, where both encoder and decoder are disentangled to facilitate learning of two sub-tasks.
- VideoGen: A Reference-Guided Latent Diffusion Approach for High Definition Text-to-Video Generation86
VideoGen sets a new state-of-the-art in text-to-video generation in terms of both qualitative and quantitative evaluation, and uses the first frame of a ground-truth video as the reference image for training the cascaded latent diffusion module.
- ViSTA: Vision and Scene Text Aggregation for Cross-Modal Retrieval86
A novel fusion token based transformer aggregation approach to exchange the necessary scene text information only through the fusion token and concentrate on the most important features in each modality to tackle the modality missing problem of scene text.
- OpenHumanVid: A Large-Scale High-Quality Dataset for Enhancing Human-Centric Video Generation84
This work introduces OpenHumanVid, a large-scale and high-quality humancentric video dataset characterized by precise and detailed captions that encompass both human appearance and motion states, along with supplementary human motion conditions, including skeleton sequences and speech audio.
- Unified Pre-training with Pseudo Texts for Text-To-Image Person Re-identification83
A new unified pre-training pipeline (UniPT) designed specifically for the T2I-ReID task is presented, in which pseudo-textual descriptions of images are automatically generated by the CLIP paradigm using a divide-conquer-combine strategy.
- Global Versus Localized Generative Adversarial Nets83
The LGAN can be applied to train a robust classifier that prefers locally consistent classification decisions on the manifold, and the resultant regularizer is closely related with the Laplace-Beltrami operator.
- Query-driven iterated neighborhood graph search for large scale indexing83
This paper presents a criterion to check if the local search over a neighborhood graph arrives at the local solution, and follows the iterated local search (ILS) strategy, widely-used in combinatorial optimization, to find a solution beyond a local optimum.
- 82
- Interactive Image Search by Color Map82
This article presents an interactive image search system, image search by color map, which can be applied to, but not limited to, enhance text-based image search, and proposes a simple but effective scheme to mine the latent search intention from the user’s input, and exploit the dominant color filter strategy to make the system more efficient.
- 81
- 80
- Finding image exemplars using fast sparse affinity propagation78
A novel approach to organize image search results obtained from state-of-the-art image search engines by adopting the idea of affinity propagation and design a fast sparse affinity propagation algorithm to find exemplars that best represent the imagesearch results.
- 77
- Beyond Attentive Tokens: Incorporating Token Importance and Diversity for Efficient Vision Transformers76
An efficient token decoupling and merging method that can jointly consider the token importance and diversity for token pruning and obtains a promising trade-off between model complexity and classification accuracy is proposed.
- Rating the Crisis of Online Public Opinion Using a Multi-Level Index System75
A method based on the multi-level evaluation index system and grey correlation analysis that can objectively evaluate the emotional tendency of Internet users and rate the crisis in different dissemination stages of online public opinion is proposed.
- Trinary-Projection Trees for Approximate Nearest Neighbor Search75
This work presents a coordinate-wise enumeration algorithm to find the principal trinary-projection direction and provides an extension using multiple randomized trees for improved performance on large-scale local patch indexing and similar image search.
- Discriminative Sketch‐based 3D Model Retrieval via Robust Shape Matching75
Critical acceleration methods to achieve interactive performance are devised: precomputing kNN graphs that record transformations between neighboring contour images and enable fast online shape alignment; pruning sampling and shape registration strategically and hierarchically; and parallelizing shape matching on multi‐core platforms or GPUs.
- Supervised Quantization for Similarity Search74
This paper presents a compact coding approach, supervised quantization, that simultaneously learns feature selection that linearly transforms the database points into a low-dimensional discriminative subspace, and quantizes the data points in the transformed space.
- 73
- Human Pose Estimation Using Global and Local Normalization71
A two-stage normalization scheme, human body normalization and limb normalization, is presented to make the distribution of the relative joint locations compact, resulting in easier learning of convolutional spatial models and more accurate pose estimation.
- Deep Convolutional Neural Networks with Merge-and-Run Mappings70
This work shows that the merge-and-run mapping is a linear idempotent function in which the transformation matrix is idem Potent, and thus improves information flow, making training easy.
- 70
- 69
- 68
- 68
- Ambiguity-Resistant Semi-Supervised Learning for Dense Object Detection67
A Ambiguity-Resistant Semi-supervised Learning (ARSL) for one-stage detectors is proposed that effectively mitigates the ambiguities and achieves state-of-the-art SSOD performance on MS COCO and PASCAL VOC.
- 67
- 65
- Semi-Supervised Classification with Universum64
This paper proposes a graph based method to make use of the Universum data to help depict the prior information for possible classifiers and shows that the proposed method can obtain superior performances over conventional supervised and semi-supervised methods.
- A Diffusion and Clustering-Based Approach for Finding Coherent Motions and Understanding Crowd Scenes63
This paper addresses the problem of detecting coherent motions in crowd scenes and presents its two applications in crowd scene understanding: semantic region detection and recurrent activity mining and introduces a cluster-and-merge process.
- Personalized video recommendation through tripartite graph propagation63
This paper develops an iterative propagation scheme over the tripartite graph to compute the preference information of each user and demonstrates that the proposed method outperforms existing state-of-the-art approaches, co-views and random walks on the user-video bipartite graph.
- Learning Correspondence Structures for Person Re-Identification62
This paper addresses the problem of handling spatial misalignments due to camera-view changes or human-pose variations in person re-identification by introducing a boosting-based approach to learn a correspondence structure, which indicates the patchwise matching probabilities between images from a target camera pair and introduces a global constraint-based matching process.
- 62
- Collaborative Quantization for Cross-Modal Similarity Search61
This paper presents a compact coding solution for efficient search, with a focus on the quantization approach which has already shown the superior performance over the hashing solutions in the single-modal similarity search.
- CAPE: Camera View Position Embedding for Multi-View 3D Object Detection60
This paper proposes a novel method based on CAmera view Position Embedding, called CAPE, which form the 3D position embed-dings under the local camera-view coordinate system instead of the global coordinate system, such that3D position em-bedding is free of encoding camera extrinsic parameters.
- 60
- No Other Representation Component Is Needed: Diffusion Transformers Can Provide Representation Guidance by Themselves59
This work proposes SelfRepresentation Alignment Alignment (SRA), a simple yet effective method that obtains representation guidance using the internal representations of learned diffusion transformer to progressively enhance the overall representation learning during only the training process.
- Instance-Specific and Model-Adaptive Supervision for Semi-Supervised Semantic Segmentation58
iMAS learns from unlabeled instances progressively by weighing their corresponding consistency losses based on the evaluated hardness and dynamically adjusts the augmentation for each instance such that the distortion degree of augmented instances is adapted to the model's generalization capability across the training course.
- Group Pose: A Simple Baseline for End-to-End Multi-person Pose Estimation58
The approach without human box supervision is superior to previous methods with complex decoders, and even is slightly better than ED-Pose that uses human box supervision.
- StrucTexTv2: Masked Visual-Textual Prediction for Document Image Pre-training58
StrucTexTv2 is an effective document image pre-training framework that models image-only input and potentially deals with more application scenarios free from OCR pre-processing, and achieves competitive or new state-of-the-art performance in various downstream tasks such as image classification, layout analysis, table structure recognition, document OCR, and information extraction.
- Group DETR v2: Strong Object Detector with Encoder-Decoder Pretraining57
Group DETR v2 achieves $\textbf{64.5}$ mAP on COCO test-dev, and establishes a new SoTA on the C OCO leaderboard.
- 57
- Image search by concept map57
A new image search interface is proposed to enable users to formulate a query, called concept map, by intuitively typing textual queries in a blank canvas to indicate the desired spatial positions of the concepts.
- Recommending Flickr groups with social topic model56
This paper presents a probabilistic latent topic model to model Flickr groups in an integrated framework, expecting to jointly discover the latent interests for users and groups and simultaneously learn the recommendation function.
- MaskOCR: Text Recognition with Masked Encoder-Decoder Pretraining54
A novel approach MaskOCR is proposed to unify vision and language pre-training in the classical encoder-decoder recognition framework, and enhance the language modeling capability of the sequence decoder using a proposed masked image-language modeling scheme.
- GPT4Vis: What Can GPT-4 Do for Zero-shot Visual Recognition?53
Evaluation of GPT-4's linguistic and visual capabilities in zero-shot visual recognition tasks shows that GPT-4, enhanced with rich linguistic descriptions, significantly improves zero-shot recognition, offering an average top-1 accuracy increase of 7% across all datasets.
- PSVT: End-to-End Multi-Person 3D Pose and Shape Estimation with Progressive Video Transformers53
A new end-to-end multi-person 3D Pose and Shape estimation framework with progressive Video Transformer, termed PSVT, and a novel pose-guided attention (PGA) for shape decoder to better predict shape parameters are proposed to improve performance.
- Group Re-Identification53
Multi-grain representations are computed to characterize the appearance and spatial features of multi-grain objects and evaluate the importance weight of each object for group Re-ID, so as to handle the interferences from group dynamics.
- 53
- A transductive multi-label learning approach for video concept detection52
This paper presents an effective transductive multi-label classification approach that simultaneously models the labeling consistency between the visually similar videos and the multi- label interdependence for each video.
- ByteTrackV2: 2D and 3D Multi-Object Tracking by Associating Every Detection Box51
A complementary motion prediction strategy that incorporates the detected velocities with a Kalman filter to address the problem of abrupt motion and short-term disappearing is proposed.
- InterActive: Inter-Layer Activeness Propagation51
InterActive is presented, a novel algorithm which computes the activeness of neurons and network connections, and achieves state-of-the-art classification performance on a wide range of image datasets.
- MonoFormer: One Transformer for Both Diffusion and Autoregression50
Experimental results show that the proposed share one transformer for both autoregression and diffusion achieves comparable image generation performance to current state-of-the-art methods as well as maintains the text generation capability.
- Group Reidentification with Multigrained Matching and Integration50
This paper introduces multigrained representations which can be extracted via the development of two separate schemes, that is, one with handcrafted descriptors and another with deep neural networks, and is further equipped with importance weights which capture variations in intragroup dynamics.
- 50
- 50
- 49
- 49
- Finding Coherent Motions and Semantic Regions in Crowd Scenes: A Diffusion and Clustering Approach48
A coarse-to-fine thermal-diffusion-based approach to detect coherent motions in crowd scenes and subsequently constructing semantic regions for activity recognition, and introduces a two-step clustering process to construct stable semantic regions from the extracted time-varying coherent motions.
- 47
- 46
- Boosting Few-shot Action Recognition with Graph-guided Hybrid Matching45
GgHM, a new framework with Graph-guided Hybrid Matching, which learns task-oriented features by the guidance of a graph neural network during class prototype construction, optimizing the intra- and inter-class feature correlation explicitly.
- 45
- Masked Lip-Sync Prediction by Audio-Visual Contextual Exploitation in Transformers45
This work proposes the Audio-Visual Context-Aware Transformer (AV-CAT) framework, which produces accurate lip-sync with photo-realistic quality by predicting the masked mouth shapes and proposes a convolution-Transformer hybrid backbone and design an attention-based fusion strategy for filling the masked parts.
- Transferring Vision-Language Models for Visual Recognition: A Classifier Perspective44
This study revisits the role of the linear classifier in the vanilla transfer learning framework, and proposes a new paradigm where the parameters of the classifier are initialized with semantic targets from the textual encoder and remain fixed during optimization.
- 44
- 44
- 43
- Unsupervised Pre-Training With Language-Vision Prompts for Low-Data Instance Segmentation42
A new method, Unsupervised Pre-training with Language-Vision Prompts (UPLVP), which improves QEIS models’ instance segmentation by bringing language-vision prompts to queries/kernels, and can converge faster and perform better than CNN-based models in low-data regimes.
- Conditional DETR V2: Efficient Detection Transformer with Box Queries42
Inspired by Conditional DETR, an improved DETR with fast training convergence, the object query is reformulate into the format of the box query that is a composition of the embeddings of the reference point and the transformation of thebox with respect to thereference point.
- Bottom-up Higher-Resolution Networks for Multi-Person Pose Estimation41
Higher-Resolution Network (HigherHRNet) is proposed, which is a simple extension of the High-Res resolution Network (HRNet), which generates higher-resolution feature maps by deconvolving the high- resolution feature maps outputted by HRNet, which are spatially more accurate for small and medium persons.
- Face recognition using spectral features41
A new unsupervised nonlinear feature extraction method called spectral feature analysis (SFA) is proposed, which can extract discriminatory information from the data, and it is shown that linear discriminant analysis can be subsumed under the SFA framework.
- 40
- 39
- Fast Neighborhood Graph Search Using Cartesian Concatenation39
Experimental results on searching over large scale datasets (SIFT, GISTand HOG) show that the proposed new data structure for approximate nearest neighbor search outperforms state-of-the-art ANN search algorithms in terms of efficiency and accuracy.
- 38
- 38
- Fixed-Point Model For Structured Labeling37
An algorithm with a new perspective on layered models; it aims to find a fixed-point function with the structured labels being both the output and the input, which alleviates the burden in learning multiple/different classifiers in different layers.
- Document clustering with universum37
A novel document clustering technique -- Document Clustering with Universum, which utilizes the Universum examples to improve the clustering performance and an extensive set of experiments are conducted to demonstrate the effectiveness and efficiency of the proposed algorithm.
- CPCM: Contextual Point Cloud Modeling for Weakly-supervised Point Cloud Semantic Segmentation36
A simple yet effective Contextual Point Cloud Modeling (CPCM) method that consists of two parts: a region-wise masking (Region-Mask) strategy and a contextual masked training (CMT) method that disentangles the learning of supervised segmentation and unsupervised masked context prediction for effectively learning the very limited labeled points and mass unlabeled points, respectively.
- On the Connection of Deep Fusion to Ensembling.36
A new deeply-fused network is developed that combines two networks in a merge-and-run fusion manner that is less deep than a ResNet but yields an ensemble of the same number of more-capable component networks, thus improving the classification accuracy.
- Self-Paced Cross-Modal Subspace Matching36
Experimental results on four multimodal databases show that SCSM outperforms state-of-the-art cross-modal subspace learning methods.
- Multi-Domain Incremental Learning for Face Presentation Attack Detection35
An adaptive domain-specific experts (ADE) framework based on the vision transformer to preserve the discriminability of previous domains and an asymmetric classifier is designed to keep the output distribution of different classifiers consistent, thereby improving the generalization ability.
- Towards Reversal-Invariant Image Representation35
A reversal-invariant version of SIFT descriptor named Max-SIFT, a generalized RIDE algorithm which can be applied to a large family of local descriptors which reveals consistent accuracy gain on various image classification tasks, including scene understanding, fine-grained object recognition, and large-scale visual recognition.
- Binary Optimized Hashing35
Binary optimized hashing (BOH) is proposed, in which it is proved that if the loss function is Lipschitz continuous, the binary optimization problem can be relaxed to a bound-constrained continuous optimization problem.
- 35
- VDG: Vision-Only Dynamic Gaussian for Driving Simulation34
Vision-only Dynamic Gaussian (VDG), a novel method that, for the first time, integrates self-supervised visual odometry (VO) into a pose-free dynamic Gaussian splatting framework, enabling precise static-dynamic decomposition and modeling of dynamic objects via dynamic Gaussians.
- Unified Frequency-Assisted Transformer Framework for Detecting and Grounding Multi-modal Manipulation34
The Unified Frequency-Assisted transFormer framework, named UFAFormer, is presented, which comprises two symmetric cross-modal interaction modules responsible for gathering modality-specific forgery information, along with a fusing interaction module for aggregation of both modalities.
- 34
- Image tag refinement by regularized latent Dirichlet allocation34
This paper proposes to do tag refinement from the angle of topic modeling and presents a novel graphical model, regularized latent Dirichlet allocation (rLDA), where tag similarity and tag relevance are jointly estimated in an iterative manner, so that they can benefit from each other.
- Image-Based Modeling by Joint Segmentation34
To structure the available 3D points and registered 2D image information, it is argued that a joint segmentation of both 3D and 2D is the fundamental stage for the subsequent modeling, and a probabilistic framework for the joint segmentations is proposed.
- Graph based image segmentation34
The prior model for the pairwise graph approaches which is defined from multiple cues, a hyper graph based method which models multiple wise relations among the data points, and a tree structured graphbased method which leads to an efficient and effective solution to the normalized cuts criterion are developed.
- Kernel Trick Embedded Gaussian Mixture Model34
This paper presents a kernel trick embedded Gaussian Mixture Model (GMM), called kernel GMM, which can solve problems in probabilistic framework and tackle nonlinear problems better than the traditional GMM.
- 33
- 32
- Few-Shot Head Swapping in the Wild32
This paper presents the Head Swapper (HeSer), which achieves few-shot head swapping in the wild through two delicately de-signed modules, and demonstrates that the proposed method produces superior head swapping results on a variety of scenes.
- 32
- 31
- 31
- Random Shifting for CNN: a Solution to Reduce Information Loss in Down-Sampling Layers31
A novel random strategy is proposed to alleviate problems of poor generalization ability and information loss in deep convolutional neural networks by embedding random shifting in the down-sampling layers during the training process.
- JIGSAW31
This paper presents an innovative application for mobile phone users to facilitate their visual search experience by taking advantage of smart phone functionalities such as multi-modal and multi-touch interactions, and demonstrates that it is an effective complement to existing mobile visual search applications.
- CAE v2: Context Autoencoder with CLIP Target30
This work studies two critical elements in MIM, i.e., the supervision position and the mask ratio, and reveals two interesting perspectives, relying on a developed simple pipeline, context autodecoder with CLIP target (CAE v2).
- 30
- BEVSpread: Spread Voxel Pooling for Bird's-Eye-View Representation in Vision-Based Roadside 3D Object Detection29
Instead of bringing the image features contained in a frustum point to a single BEV grid, BEVSpread considers each frustum point as a source and spreads the image features to the surrounding BEV grids with adaptive weights to achieve superior prop- agation performance.
- GVA: Reconstructing Vivid 3D Gaussian Avatars from Monocular Videos28
A pose refinement technique is introduced to improve hand and foot pose accuracy by aligning normal maps and silhouettes and addressing the problems of unbalanced aggregation and initialization bias that previously diminished the quality of 3D Gaussian avatars through a novel surface-guided re-initialization method.
- TRUST: An Accurate and End-to-End Table structure Recognizer Using Splitting-based Transformers28
An accurate and end-to-end transformer-based table structure recognition method, referred to as TRUST, that surpasses the previous methods by a large margin and achieves new state-of-the-art results.
- 28
- Joint salient object detection and existence prediction28
This paper proposes a supervised learning approach for jointly addressing the salient object detection and existence prediction problems and adopts the structural SVM framework, which forms the two problems jointly in a single integrated objective function.
- Personalized Video Recommendation through Graph Propagation28
This article addresses the problem of providing personalized video recommendation for users by combining the clicks and queries information to build a tripartite graph and proposing an iterative propagation algorithm over the enhanced graph to compute the preference information of each user.
- MA-FSAR: Multimodal Adaptation of CLIP for few-shot action recognition27
MA-FSAR is introduced, a framework that employs the Parameter-Efficient Fine-Tuning (PEFT) technique to enhance the CLIP visual encoder in terms of action-related temporal and semantic representations and proposes a prototype-level text-guided construction module to further enrich the temporal and semantic characteristics of video prototypes.
- Decoupled Pseudo-Labeling for Semi-Supervised Monocular 3D Object Detection27
This work introduces a novel decoupled pseudo-labeling (DPL) approach for SSM30D, and presents a Depth Gradient Projection module to mitigate optimization conflicts caused by noisy depth supervision of pseudo-labels, effectively decoupling the depth gradient and re-moving conflicting gradients.
- A Survey of Multimodal Controllable Diffusion Models27
This survey contextualize recent results, provide the dedicated treatment of controllable diffusion model generation, and outline limitations and future directions, and provides a detailed taxonomy defining controlled generation for diffusion models.
- Composite Quantization27
This paper studies the compact coding approach to approximate nearest neighbor search and theoretically justifies the equivalence between near-orthogonal composite quantization and minimizing an upper bound of a function formed by jointly considering the quantization error and the search cost according to a generalized triangle inequality.
- 26
- GridFormer: Towards Accurate Table Structure Recognition via Grid Prediction26
GridFormer, a novel approach for interpreting unconstrained table structures by predicting the vertex and edge of a grid by introducing a DETR-style table structure recognizer to efficiently predict this multi-objective information of the grid in a single shot.
- Interactive browsing via diversified visual summarization for image search results26
A novel search result summarization approach is introduced and an interactive browsing scheme, based on a tree structure for organizing the images obtained from the summarization approaches, is proposed to enable users to intuitively and conveniently browse the image search results.
- Transductive multi-label learning for video concept detection26
This paper proposes an effective transductive multi-label classification approach, which simultaneously models the labeling consistency between the visually similar videos and the multi- labels interdependence for each video in an integrated framework and compares the performance between the proposed approach and several representative transductives single-label and supervised multi- label classification approaches.
- 25
- 25
- What Can Simple Arithmetic Operations Do for Temporal Modeling?24
This work explores the potential of four simple arithmetic operations for temporal modeling by computing addition, subtraction, multiplication, and division between pairs of extracted frame features to benefit the original temporal-irrespective domain.
- 24
- 23
- 23
- Hybrid image summarization23
A novel approach, called Homogeneous and Heterogeneous Message Propagation (H2MP), which extends affinity propagation that only works over homogeneous relations to heterogeneous relations and the summary obtained by this approach is both visually and semantically satisfactory.
- Distillation-Guided Residual Learning for Binary Convolutional Neural Networks22
This work enforces BCNN to produce similar intermediate feature maps with the ones of FCNN to minimize the performance gap, and updates the binary convolutional block architecture to facilitate the optimization of blockwise distillation loss.
- 22
- Interactive image search by 2D semantic map22
A novel interactive image search system, image search by 2D semantic map, that enables users to indicate what semantic concepts are expected to appear and even how these concepts are spatially distributed in the desired images is presented.
- Towards Lightweight Super-Resolution With Dual Regression Learning21
A dual regression compression method is proposed to reduce model redundancy in both layer-level and channel-level based on channel pruning to alleviate the first challenge and show the effectiveness of the method in obtaining accurate and efficient SR models.
- 21
- 21
- 21
- TALK-Act: Enhance Textural-Awareness for 2D Speaking Avatar Reenactment with Diffusion Model20
The Motion-Enhanced Textural-Aware ModeLing for SpeaKing Avatar Reenactment (TALK-Act) framework, which enables high-fidelity avatar reenactment from only short footage of monocular video, and carefully construct 2D and 3D structural information as intermediate guidance.
- Revisiting MLLMs: An In-Depth Analysis of Image Classification Abilities20
This paper thoroughly revisiting the MLLMs with an in-depth analysis of image classification, revealing that the most recent MLLMs can match or even outperform CLIP-style vision-language models on several datasets, challenging the previous assumption that MLLMs are bad at image classification.
- 20
- 20
- Joint Multilabel Classification With Community-Aware Label Graph Learning20
This work proposes a multilabel classification framework based on a joint learning approach called label graph learning (LGL) driven weighted Support Vector Machine (SVM) which explicitly models the inter-label correlations by LGL, which is jointly optimized with multILabel classification in a unified learning scheme.
- Surfel-Based Gaussian Inverse Rendering for Fast and Relightable Dynamic Human Reconstruction From Monocular Videos19
SGIA (Surfel-based Gaussian Inverse Avatar) is presented, which introduces efficient training and rendering for relightable dynamic human reconstruction and significantly enhances the realistic relighting of dynamic human avatars, providing a substantial speed advantage.
- Generating Action-conditioned Prompts for Open-vocabulary Video Action Recognition19
It is argued that augmenting text embeddings with human prior knowledge is pivotal for open-vocabulary video action recognition, and innovatively blend video models with Large Language Models (LLMs) to devise Action-conditioned Prompts.
- 19
- Optimized Distances for Binary Code Ranking19
Extensive experimental results show superior performance of the proposed approach over state-of-the-art methods on three real-world high-dimensional datasets for binary code ranking.
- MSRA atT TRECVID 2008: High-Level Feature Extraction and Automatic Search.19
This paper describes the MSRA experiments for TRECVID 2008, which representatively investigated the benefit of global and local low-level features by a variety of learning-based methods, including supervised and semi-supervised learning algorithms.
- Object Detection in Videos by Short and Long Range Object Linking18
This work proposes a set of candidate spatio-temporal cuboids, each of which serves as a container associating the object across short range frames, for a short video segment and links the tubelets across temporally-overlapping short segments over the whole video in order to boost the classification scores for positive detections by aggregating the scores in the linked tubelets.
- Exemplar-Guided Similarity Learning on Polynomial Kernel Feature Map for Person Re-identification18
A novel organization of polynomial kernel feature map in a high dimensional feature space to break down the variability of positive person pairs and forms a unified learning problem including a relaxed loss term as well as two kinds of regularization strategies particularly designed for the feature map.
- Dual Low-Rank Pursuit: Learning Salient Features for Saliency Detection18
This paper considers a specific saliency detection problem of predicting human eye fixations when they freely view natural images, and proposes a novel dual low-rank pursuit (DLRP) method that learns saliency-aware feature transformations by utilizing available supervision information and constructs discriminative bases for effectively detecting human fixation points under the popular low- rank and sparsity-pursuit framework.
- Fast Nearest Neighbor Search in the Hamming Space18
This paper generalizes the algorithm to the Hamming space with an alternative version of k-means clustering and achieves competitive performance compared to the state-of-the-art methods, i.e., MIH and FLANN, in the aspects of search precision, accessed data volume and average querying time.
- Incorporating visual adjectives for image classification18
This paper proposes a simple framework to integrate various types of adjectives, i.e., color, shape and location, for more powerful image representation, and shows that this approach achieves superior classification accuracy with reasonable computational overheads.
- 18
- Transductive 3D Shape Segmentation using Sparse Reconstruction18
A transductive shape segmentation algorithm, which can transfer prior segmentation results in database to new shapes without explicitly specification of prior category information, and can generate high quality segmentation and semantic labeling results in the Princeton segmentation benchmark.
- 17
- M2-CLIP: A Multimodal, Multi-task Adapting Framework for Video Action Recognition17
This work designs a multi-task decoder with a rich set of supervisory signals, including the original contrastive learning head, a cross-modal classification head, a cross-modal masked language modeling head, and a visual classification head that adeptly satisfies the need for strong supervised performance within a multimodal framework.
- Learning Structure-Guided Diffusion Model for 2D Human Pose Estimation17
A new scheme that formulates 2D HPE as a keypoints heatmaps generation problem from noised heatmaps from a progressive denoising way, and further explores improving the performance of DiffusionPose with conditions from human structural information.
- 17
- 17
- 17
- Feature Incay for Representation Regularization17
This work proposes feature incay to also regularize representation learning, which favors feature vectors with large norm when the samples can be correctly classified, and encourages intra-class compactness along the directions of weight vectors.
- Multi-Modal 3D Object Detection by Box Matching16
A novel Fusion network by Box Matching (FBMNet) for multi-modal 3D detection is proposed, which provides an alternative way for cross-modal feature alignment by learning the correspondence at the bounding box level to free up the dependency of calibration during inference.
- StrucTexTv3: An Efficient Vision-Language Model for Text-rich Image Perception, Comprehension, and Beyond16
The StrucTexTv3 model, featuring efficient structural design, outstanding performance, and broad adaptability, offers robust support for diverse intelligent application tasks involving text-rich images, thus exhibiting immense potential for widespread application.
- NeRF-Loc: Transformer-Based Object Localization Within Neural Radiance Fields16
A transformer-based framework, NeRF-Loc, to extract 3D bounding boxes of objects in NeRF scenes, which is better than the baselines and compared with conventional RGB(-D) based methods that take rendered RGB images and depths from NeRFs as inputs.
- Paint and Distill16
This work proposes a novel semantic passing framework, named SPNet, to boost the performance of existing lidar-based 3D detection models with the guidance of rich context painting, with no extra computation cost during inference.
- 16
- Image search by graph-based label propagation with image representation from DNN16
This paper proposes a graph-based label propagation algorithm that employs neighborhood graph search to find the nearest neighbors on an image similarity graph built up with visual representations from deep neural networks and further aggregates their clicked queries/click counts to get the labels of the new image.
- 16
- 15
- MGMapNet: Multi-Granularity Representation Learning for End-to-End Vectorized HD Map Construction15
A simple yet efficient framework to model map element with a multi-granularity representation, integrating both coarse-grained instance-level and fine-grained point-level queries, and a Point Instance Interaction module to encourage information exchange between instance-level and point-level queries.
- 15
- 15
- Disentangled Representation Learning With Transmitted Information Bottleneck15
This work proposes DisTIB (Transmitted Information Bottleneck for Disentangled representation learning), a novel objective that navigates the balance between information compression and preservation and theoretically proves that DisTIB can achieve optimal disentanglement.
- Browse-to-Search15
This work develops a novel system to explore an effective and efficient way for allowing users to seamlessly switch between browse and search processes, and naturally complete visual-based exploratory search tasks, and provides users with more robust results to satisfy their exploring experience.
- Contextual image search15
The key idea is to use the search-to-annotation technique and the contextual textual query mining scheme to determine the corresponding textual query, to finally get semantically similar search results.
- 14
- 14
- 14
- 14
- Detecting Deepfake by Creating Spatio-Temporal Regularity Disruption14
This work proposes to disrupt a real video through a Pseudo-fake Generator and create a wide range of pseudo-fake videos for training to boost the generalization of deepfake detection by distinguishing the regularity disruption that does not appear in real videos.
- 14
- 14
- TopoSD: Topology-Enhanced Lane Segment Perception with SDMap Prior13
This work proposes to train the perception model to "see" standard definition maps (SDMaps), and incorporates such complementary features as prior information to improve the bird's eye view (BEV) feature for lane geometry and topology decoding.
- Graph Contrastive Learning for Skeleton-based Action Recognition13
A graph contrastive learning framework for skeletonbased action recognition (SkeletonGCL) to explore the global context across all sequences and establishes a new training paradigm, and it can be seamlessly incorporated into current GCNs.
- It Takes Two: Masked Appearance-Motion Modeling for Self-supervised Video Transformer Pre-training13
This work explicitly investigate motion cues in videos as extra prediction target and proposes the Masked Appearance-Motion Modeling (MAM2) framework, which designs an encoder-regressor-decoder pipeline for this task.
- Self-Guided Hard Negative Generation for Unsupervised Person Re-Identification13
A joint framework is developed which incorporates a hard negative generation network (HNGN) and a re-ID network which is alternately trained in an adversarial manner to improve each other, where the reID network guides H NGN to generate challenging data and HNGN enforces the re-IDs to enhance discrimination ability.
- 13
- Special issue on intelligent urban computing with big data13
The intelligent urban computing, which holistically exploits the big data in big cities to improve the urban environment, human life quality, and city operation systems, has obtained massive attention in research and industrial fields.
- Learning Multiple Probabilistic Decisions from Latent World Model in Autonomous Driving12
LatentDriver, a framework models the environment's next states and the ego vehicle's possible actions as a mixture distribution, from which a deterministic control signal is then derived, and achieves expert-level performance on the recently released closed-loop benchmark Waymax.
- AudCast: Audio-Driven Human Video Generation by Cascaded Diffusion Transformers12
AudCast is proposed, a generalized audio-driven human video generation framework adopting a cascade Diffusion-Transformers (DiTs) paradigm, which synthesizes holistic human videos based on a reference image and a given audio to directly drive the movements of any human body with vivid gesture dynamics.
- 12
- 12
- 12
- Probabilistic tangent subspace: a unified view12
This paper extends TD to more general pattern classification tasks, and proposes three probabilistic subspace models to encode the variations: the linear sub space, nonlinear subspace, and manifold sub space models.
- MagicGeo: Training-Free Text-Guided Geometric Diagram Generation11
The proposed MagicGeo framework formsulates the diagram generation process as a coordinate optimization problem, ensuring geometric correctness through a formal language solver, and then employs coordinate-aware generation, which provides a scalable, accurate solution for automated diagram generation.
- CSDG-FAS: Closed-Space Domain Generalization for Face Anti-spoofing11
This work reframe domain generalization-based FAS as an anomaly detection problem, positing that real faces tend to aggregate within a compact, closed space, whereas spoof faces exhibit a preference for dispersion within an open space.
- FullAnno: A Data Engine for Enhancing Image Comprehension of MLLMs11
The FullAnno system is devised, which is a data engine that can generate large-scale, high-quality, and fine-grained image annotations consisting of the category and position of objects, region descriptions, text information, as well as image dense captions.
- TexRO: Generating Delicate Textures of 3D Models by Recursive Optimization11
This paper presents TexRO, a novel method for generating delicate textures of a known 3D mesh by optimizing its UV texture with an optimal viewpoint selection strategy, that finds the most miniature set of viewpoints covering all the faces of a mesh.
- Gradient-based Sampling for Class Imbalanced Semi-supervised Object Detection11
This work comprehensively study the class imbalance problem for SSOD under more challenging scenarios, thus forming the first experimental setting for class imbalanced SSOD (CI-SSOD), and proposes a simple yet effective gradient-based sampling framework that tackles the class imbalance problem from the perspective of two types of confirmation biases.
- Efficient Video Portrait Reenactment via Grid-based Codebook11
The Video Portrait via Grid-based Codebook (VPGC) framework is proposed, which achieves efficient and high-fidelity portrait modeling and designs a Prior-Guided Driving Module to predict reliable features from the driving signals, which can be later decoded back to high-quality video portraits by querying the codebook.
- Content-aware convolutional neural networks11
A Content-aware Convolution (CAC) is proposed that automatically detects the smooth windows and applies a 1 ×1 convolutional kernel to replace the original large kernel, and is able to effectively avoid the redundant computation on similar pixels.
- Geometric Neural Phrase Pooling: Modeling the Spatial Co-occurrence of Neurons11
The Geometric Neural Phrase Pooling algorithm is proposed, which acts as a new type of hidden layer, which punishes the isolated neuron responses after convolution, and can be inserted into a CNN model with little extra computational overhead.
- 11
- Deforming Videos to Masks: Flow Matching for Referring Video Segmentation10
FlowRVS, a novel framework that reconceptualizes RVOS as a conditional continuous flow problem, and reformulates the task by learning a direct, language-guided deformation from a video's holistic representation to its target mask, achieving new state-of-the-art results across all major RVOS benchmarks.
- TryOn-Adapter: Efficient Fine-Grained Clothing Identity Adaptation for High-Fidelity Virtual Try-On10
This work decouple clothing identity into fine-grained factors: style for color and category information, texture for high-frequency details, and structure for smooth spatial adaptive transformation and introduces the training-free T-RePaint strategy to further enhance clothing identity preservation while maintaining the realistic try-on effect during the inference.
- XLD: A Cross-Lane Dataset for Benchmarking Novel Driving View Synthesis10
This paper presents a synthetic dataset for novel driving view synthesis evaluation, which is specifically designed for autonomous driving simulations and establishes the first realistic benchmark for evaluating existing NVS approaches under frontonly and multicamera settings.
- 10
- Understanding Self-Supervised Pretraining with Part-Aware Representation Learning10
The empirically compare the off-the-shelf encoders pretrained with several representative methods on object- level recognition and part-level recognition and show that the fully-supervised model outperforms self-super supervised models for object-level Recognition and the combination of contrastive learning and masked image modeling further improves the performance.
- 10
- 10
- Improving Person Re-Identification With Iterative Impression Aggregation10
This work proposes a simple attentional aggregation formulation to instantiate the problem of person re-identification (re-ID), where the representation of a query image is iteratively updated with new information from the candidates in the gallery, and demonstrates competitive performance on standard benchmarks including CUHK03, Market-1501 and DukeMTMC.
- Beyond Intra-modality Discrepancy: A Comprehensive Survey of Heterogeneous Person Re-identification.10
The state-of-the-art methods comprehensively with respect to four main application scenarios -- low-resolution, infrared, sketch and text are reviewed.
- The Study of MDM2 rs937283 Variant and Cancer Susceptibility in a Central Chinese Population10
The case–control analysis revealed that rs937283 was associated with the susceptibility to breast and liver cancer, but not cervical, colon, or rectal cancer, and this significant association remained in Asian (Chinese) population,but not in Caucasian population.
- Multi-Timescale Collaborative Tracking10
The multi-timescale collaborative tracker for single object tracking combines different types of "forces", namely attraction, repulsion and support, to take advantage of their complementary strengths and outperforming current state-of-the-art methods.
- Color filter for image search10
A color filter scheme for image search is shown, which enables users to find images containing objects or scenes with their interested color, one of the most crucial cues in describing visual contents.
- EasyChauffeur: A Baseline Advancing Simplicity and Efficiency on Waymax9
This paper introduces EasyChauffeur, a reproducible and effective planner for both imitation learning (IL) and reinforcement learning (RL) on Waymax, a GPU-accelerated simulator, and proposes Ego-Shifting, a new evaluation setting for assessing planners' robustness.
- Mutual Prompt Leaning for Vision Language Models9
A fine-grained text prompt (FTP) that decomposes the single global image features into several finer-grained semantics and incorporates them into corresponding text prompt tokens, and a text-reorganized visual prompt (TVP) that reorganizes the text descriptions of the current image to construct the visual prompt.
- Vision Transformer with Attention Map Hallucination and FFN Compaction9
This work proposes to hallucinate half of the attention maps from the rest with much cheaper operations, which is called hallucinated-MHSA (hMHSA), and factorize its hidden-to-output projection matrix and leverage the re-parameterization technique to strengthen its capability, making it compact-FFN (cFFN).
- Robust Video Portrait Reenactment via Personalized Representation Quantization9
The Video Portrait via Non-local Quantization Modeling (VPNQ) framework is proposed, which produces pose- and disturbance-robust reenactable video portraits and a personalized one can be trained to preserve desired position-invariant local details better.
- A simple algorithm for the implementation of second-order-polynomial based peak-tracking methods9
A simple algorithm is proposed for the implementation of second-order-polynomial-based peak-tracking methods used in FBG sensors by decomposed the Moore-Penrose of the coefficient matrix into the multiplication of an upper triangular matrix and a symmetric matrix, and derived an equation nearly as simple as that of the centroid method for the determination of peak wavelength.
- 9
- 9
- PSDiff: Diffusion Model for Person Search With Iterative and Collaborative Refinement8
A novel Person Search framework based on the Diffusion model, PSDiff, which formulates the person search as a dual denoising process from noisy boxes and ReID embeddings to ground truths and discards prior pedestrian candidates generated by detectors, thereby avoiding the local optimum problem of the ReID task.
- 8
- A comprehensive study of CD44 rs 187115 variant and cancer risk in a central Chinese population8
Collectively, the CD44 rs187115 variant may be associated with the risk of cervical, lung, and liver cancer in the central Chinese population, and may be used as a potential biomarker for cancer predisposition in the Asian population.
- 8
- 8
- 8
- Visual object recognition using probabilistic kernel subspace similarity8
Experiments on synthetic data and real visual object recognition tasks show that probabilistic kernel subspace similarity-based face matching can achieve promising performance, and outperform many other current popular object recognition algorithms.
- Disentangled Noisy Correspondence Learning7
Inspired by the robustness of information bottlenecks against noise, DisNCL is introduced, a novel information-theoretic framework for feature Disentanglement in Noisy Correspondence Learning to adaptively balance the extraction of modality-invariant information (MII) and MEI with certifiable optimal cross-modal disentanglement efficacy.
- DGTR: Distributed Gaussian Turbo-Reconstruction for Sparse-View Vast Scenes7
DGTR is presented, a novel distributed framework for efficient Gaussian reconstruction for sparse-view vast scenes and novel-view synthesis in significantly reduced training times, outperforming existing approaches in both speed and scalability.
- Are Images Indistinguishable to Humans Also Indistinguishable to Classifiers?7
This work reveals that, from the perspective of neural network-based classifiers, even advanced diffusion models are still far from perfect data distribution and sheds light on the model autophagy disorder and offers insights into the use of generated data: augmenting real data with generated data is more effective than replacing it.
- ALoRE: Efficient Visual Adaptation via Aggregating Low Rank Experts7
ALoRE is proposed, a novel PETL method that reuses the hypercomplex parameterized space constructed by Kronecker product to Aggregate Low Rank Experts using a multi-branch paradigm, disentangling the learned cognitive patterns during training.
- Hybrid Network Compression via Meta-Learning7
This paper incorporates two criteria and seeks layer-wise compression by leveraging the meta-learning framework and proposes an iterative compression constraint for optimizing the compression procedure, which effectively achieves a high compression rate and maintains the original network performance.
- 7
- 7
- Training Better CNNs Requires to Rethink ReLU.7
The proportional module with N:M (N$>$M) convolution and ReLU ratio can help networks acquire the better performance, through the analysis of a simple Leaky ReLU model and the experimental results verify that the superiority of the proportional module is found.
- 7
- 7
- Scalable similar image search by joint indices7
This work presents an efficient and effective indexing algorithm, neighborhood graph index, which is suitable for millions of images, and uses it to organize joint inverted indices to search over billions of images.
- 7
- 7
- Training-Free Unsupervised Prompt for Vision-Language Models6
The proposed Training-Free Unsupervised Prompts (TFUP), which maximally preserves the inherent representation capabilities and enhances them with a residual connection to similarity-based prediction probabilities in a training-free and labeling-free manner, achieves surprising performance, even surpassing the training-base method on multiple classification datasets.
- 6
- 6
- 6
- Task-Oriented Multi-Modal Mutual Leaning for Vision-Language Models6
This work proposes a class-aware text prompt (CTP), which can effectively involve image semantics and avoid introducing extra ambiguities into different prompts, and proposes text-guided feature tuning (TFT) to make the image branch attend to class-related representation.
- 6
- 6
- 6
- 6
- Mixture Factorized Ornstein-Uhlenbeck Processes for Time-Series Forecasting6
A Mixture Factorized OU process (MFOUP) is developed that is able to capture the changing states of multiple mixed hidden factors, from which it can infer their roles in driving the movements of time series.
- Similar image search with a tiny bag-of-delegates representation6
A tiny bag-of-delegates representation that uses a small amount of delegates with a high search performance guaranteed is proposed, and an adaptive forward selection scheme to sequentially learn more and more inverted indices that are constructed based on subspace partition, e.g. using spatial partition trees.
- Color Image Segmentation: Kernel Do the Feature Space6
The scheme is that an extension of EM algorithm for GMM is proposed by involving one kernel feature extraction step, which is called K-EM, which not only speeds up kernel step, but also automatically extracts good features for clustering in a nonlinear way.
- 6
- Cosh-DiT: Co-Speech Gesture Video Synthesis via Hybrid Audio-Visual Diffusion Transformers5
Cosh-DiT, a Co-speech gesture video system with hybrid Diffusion Transformers that perform audio-to-motion and motion-to-video synthesis using discrete and continuous diffusion modeling, respectively, and designs a visual Diffusion Transformer that effectively integrates spatial and temporal contexts.
- UniLION: Towards Unified Autonomous Driving Model with Linear Group RNNs5
A unified autonomous driving model, UniLION, which efficiently handles large-scale LiDAR point clouds, high-resolution multi-view images, and even temporal sequences based on the linear group RNN operator (i.e., performs linear RNN for grouped features).
- Fusion4DAL: Offline Multi-modal 3D Object Detection for 4D Auto-labeling5
This work proposes a simple yet effective offline multi-modal 3D object detection method, named Fusion4DAL, for 4D auto-labeling with long-term multi-modal sequences, and proposes a multi-modal mixed feature fusion module (MMFF), to address the sparsity of points within objects.
- 5
- 5
- 5
- Low-rank SIFT: An affine invariant feature for place recognition5
A novel affine-invariant feature, Low-rank SIFT, which exploits the regular appearance property in man-made objects and achieves full affine invariance without needing to simulate over affine parameter space is presented.
- 5
- Dynamic Video Collage5
This demo presents a video visualization technique named Dynamic Video Collage (DVC), which aims to mimic the storytelling of the video content by dynamically presenting a series of collages.
- 5
- 4
- Pyramidal Patchification Flow for Visual Generation4
This work introduces a Pyramidal Patchification Flow approach, which operates over full latent representations other than pyramid representations, and adopts the normal denoising process without requiring the renoising trick.
- MomentSeg: Moment-Centric Sampling for Enhanced Video Pixel Understanding4
A unified framework that jointly optimizes Temporal Sentence Grounding (TSG) and RefVOS, naturally incorporating key moment grounding capability is proposed, and a Moment-Centric Sampling (MCS) strategy that densely samples informative moments while sparsely sampling non-essential frames is designed.
- Towards Unified Multi-granularity Text Detection with Interactive Attention4
Experimental results demonstrate that DAT achieves state-of-the-art performances across a variety of text-related benchmarks, including multi-oriented/arbitrarily-shaped scene text detection, document layout analysis and page detection tasks.
- 4
- Diverse Learner: Exploring Diverse Supervision for Semi-supervised Object Detection4
To maintain distinctiveness between teachers and students, the diverse learner framework consists of two paired teacher-student models with diverse supervision strategy and argues that the pseudo labels which are typically regarded as unreliable and obsoleted by many existing methods are of great value.
- 4
- 4
- 4
- How Fashion Talks: Clothing-Region-Based Gender Recognition4
This paper investigates the gender recognition problem of people in photos via clothing information other than faces in the case of insufficient face specification using category-level clothing parsing and an offline learned Mahalanobis distance metric on the middle level attributes to yield a final prediction on gender.
- Color-Structured Image Search4
A novel interactive image ranking technique, color-structured image search, by exploiting color spatial relation information, is proposed and demonstrated in image research results refinement and a convenient interactive interface is presented to allow users to specify target color structure flexibly.
- Graph Based Image Segmentation: A modern approach4
The prior model for the pairwise graph approaches which is defined from multiple cues, a hyper graph based method which models multiple wise relations among the data points, and a tree structured graphbased method which leads to an efficient and effective solution to the normalized cuts criterion are developed.
- 3
- 3
- iDiT-HOI: Inpainting-based Hand Object Interaction Reenactment via Video Diffusion Transformer3
Comprehensive evaluations demonstrate that the novel framework iDiT-HOI outperforms existing methods, particularly in challenging real-world scenes, offering enhanced realism and more seamless hand-object interactions.
- Skim then Focus: Integrating Contextual and Fine-grained Views for Repetitive Action Counting3
A dual-branch network, i.e., SkimFocusNet, working in a two-step manner, that can perform specified action counting, that is, to enable counting a particular action type by referencing an exemplary video, substantially exhibits the method’s robustness.
- TriCLIP-3D: A Unified Parameter-Efficient Framework for Tri-Modal 3D Visual Grounding based on CLIP3
A unified 2D pre-trained multi-modal network to process all three modalities (RGB images, text, and point clouds) and achieves unified feature extraction and fusion across the three modalities, enabling an end-to-end 3D visual grounding model.
- OVLW-DETR: Open-Vocabulary Light-Weighted Detection Transformer3
This report provides an end-to-end training recipe that transferring knowledge from vision-language model (VLM) to object detector with simple alignment, and demonstrates that the proposed approach is superior over existing real-time open-vocabulary detectors on standard Zero-Shot LVIS benchmark.
- 3
- 3
- Finding the Secret of CNN Parameter Layout under Strict Size Constraint3
A novel energy function is proposed to represent the CNN parameter layout, which makes it possible to model the relationship between the allocation of parameters in the convolution layers and the discriminative ability of CNN.
- Guest Editorial: Big Media Data: Understanding, Search, and Mining3
This editorial, as the fourth part of this special issue, introduces four papers on fine-grained image classification, object retrieval, and applications of social media.
- 3
- 3
- 2
- 2
- 2
- 2
- Accelerating Vision Transformers Based on Heterogeneous Attention Patterns2
An integrated compression pipeline based on observed heterogeneous attention patterns across layers and a method of global aggregation pyramid (GLAD) to reduce the number of tokens in later layers of ViTs, such as Deit.
- 2
- 2
- Decoupled Convolutions for CNNs2
This work proposes an effective approach to relax the sparsity of the filter in spatial aggregation by learning a spatial configuration, and reduce the redundancy by reducing the number of intermediate channels by decoupling the convolution along the spatial and channel domains.
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 1
- Low-Biased General Annotated Dataset Generation1
A bi-level semantic alignment loss is developed, which not only forces all generated images to be consistent with the semantic distribution of all categories belonging to the target dataset in an adversarial learning manner, but also requires each generated image to match the semantic description of its category name.
- 1
- 1
- Collaborative Position Reasoning Network for Referring Image Segmentation1
This work executes a Collaborative Position Reasoning Network (CPRN) via the proposed novel Row-and-Column interactive (RoCo) and Guided Holistic interactive (Holi) modules, the first work that explicitly focuses on position reasoning modeling.
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- Primal-Dual Group Convolutions for Deep Neural Networks1
This paper presents a simple and modularized neural network architecture, named primal-dual group convolutional neural networks (PDGCNets), and demonstrates that these networks are more efficient in using parameters and computation complexity with similar or higher accuracy.
- Building Semantic Enrichment Framework for Scientific Literature Retrieval System1
A semantic multidimensional index is built for scientific literature retrieval system with the help of semantic recognition and knowledge relationship computing and a new data organization model is designed.
- 1
- 1
- Query-driven iterated neighborhood graph search for scalable visual indexing1
This paper investigates a simple but very effective approach, neighborhood graph (NG) search, which conducts the local search by expanding neighborhoods with a best-first manner, and proposes a query-driven iterated neighborhood graph search approach to improve the performance.
- 1
- Web-scale image search by color sketch1
This demonstration presents an image search system, which enables consumers to find images on the requirement of how the colors are spatially distributed, and is the first one to scale up to Web-scale images.
- 1
- Exploring Multimodal Adapters with Multi-Task Decoders for Video Action Recognition–
A novel multimodal, multi-task CLIP adaptation framework, M-2-CLIP, which effectively balances high supervised performance with robust generalization and updates the visual adapter to TED-Adapter++, which employs a dual learnable bilinear fusion strategy to achieve this sophisticated interaction modeling.
- –
- –
- –
- –
- –
- –
- –
- –
- Assessing Model Generalization in Vicinity–
This paper proposes incorporating responses from neighboring test samples into the correctness assessment of each individual sample, and shows that applying the VRP method to existing generalization indicators, such as average confidence and effective invariance, consistently improves over these baselines both methodologically and experimentally.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Classification of ancient glassware based on Random Forest–
Through the training of random forest, the characteristic importance of each index is obtained, and the classification rule is obtained; it is determined that lead oxide is the most important index in the classification of glass types.
- RPRM as a potential preventive and therapeutic target for radiation-induced brain injury via multiple mechanisms–
RPRM knockout was found to alleviate radiation-induced brain injury in mice via targeting its multiple underlying mechanisms, suggesting that RPRM may serve as a novel potential target for the prevention and treatment of RIBI.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- On the Large-Scale Transferability of Convolutional Neural Networks–
This paper investigates the effective transfer of the CNN descriptors in generic and fine-grained classification at a large scale and provides some simple yet effective methods in constructing a competitive baseline recognition system.
- –
- –
- –
- –
- –
- –
- –
- Improved method of parameters identification and restoration of motion blurred image–
The experimental results show that the motion blurred direction can be identified effectively by the new method, and the mean square error is reduced to 68.55% compared with the old method.
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-14. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.