Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks243 from public data
- Co-occurrence Feature Learning from Skeleton Data for Action Recognition and Detection with Hierarchical Aggregation588
This paper introduces a global spatial aggregation scheme, which is able to learn superior joint co-occurrence features over local aggregation, and consistently outperforms other state-of-the-arts on action recognition and detection benchmarks like NTU RGB+D, SBU Kinect Interaction and PKU-MMD.
- Focusing Attention: Towards Accurate Text Recognition in Natural Images583
The FAN (the abbreviation of Focusing Attention Network) method is proposed that employs a focusing attention mechanism to automatically draw back the drifted attention in scene text images and substantially outperforms the existing methods.
- Skeleton-based action recognition with convolutional neural networks421
A novel convolutional neural networks (CNN) based framework for both action classification and detection of skeleton-based action recognition and a window proposal network to extract temporal segment proposals, which are further classified within the same network.
- RPVNet: A Deep and Efficient Range-Point-Voxel Fusion Network for LiDAR Point Cloud Segmentation376
This work proposes a novel range-point-voxel fusion network, namely RPVNet, and devise a deep fusion framework with multiple and mutual information interactions among these three views, and proposes a gated fusion module (termed as GFM), which can adaptively merge the three features based on concurrent inputs.
- Dynamic GCN: Context-enriched Topology Learning for Skeleton-based Action Recognition362
This paper proposes Dynamic GCN, in which a novel convolutional neural network named Context-encoding Network (CeN) is introduced to learn skeleton topology automatically, and achieves state-of-the-art performance on three large-scale benchmarks, namely NTU-RGB+D, NTU+D 120 and Skeleton-Kinetics.
- Counterfactual Samples Synthesizing for Robust Visual Question Answering356
A model-agnostic Counterfactual Samples Synthesizing (CSS) training scheme that significantly improves both visual-explainable and question-sensitive abilities of VQA models and, in return, the performance of these models is further boosted.
- Advancing Image Understanding in Poor Visibility Environments: A Collective Benchmark Study351
The UG2+ challenge Track 2 competition in IEEE CVPR 2019 is launched, aiming to evoke a comprehensive discussion and exploration about whether and how low-level vision techniques can benefit the high-level automatic visual recognition in various scenarios.
- 328
- Forward Compatible Few-Shot Class-Incremental Learning304
Virtual prototypes are assigned to squeeze the embedding of known classes and reserve for new classes, which allow the model to accept possible updates in the future and act as proxies scattered among embedding space to build a stronger classifier during inference.
- Learning Open Set Network with Discriminative Reciprocal Points294
A new concept, Reciprocal Point, which is the potential representation of the extra-class space corresponding to each known category, which can indirectly introduce the unknown information into the learner with only known classes, so as to learn more compact and discriminative representations.
- Divide-and-Assemble: Learning Block-wise Memory for Unsupervised Anomaly Detection220
By varying the granularity of division on feature maps, this work is able to modulate the reconstruction capability of the model for both normal and abnormal samples, and achieves state-of-the-art performance on the challenging MVTec AD dataset.
- Few-Shot Class-Incremental Learning by Sampling Multi-Phase Tasks211
This work proposes a new paradigm for FSCIL based on meta-learning by LearnIng Multi-phase Incremental Tasks (Limit), which synthesizes fake F SCIL tasks from the base dataset and constructs a calibration module based on transformer, which calibrates the old class classifiers and new class prototypes into the same scale and fills in the semantic gap.
- All You Need is Beyond a Good Init: Exploring Better Solution for Training Extremely Deep Convolutional Neural Networks with Orthonormality and Modulation198
A variant of regularizer which utilizes orthonormality among different filter banks can alleviate the problem of gradient vanishing or exploding phenomenon, and a backward error modulation mechanism based on the quasi-isometry assumption between two consecutive parametric layers is designed.
- Small-Scale Pedestrian Detection Based on Topological Line Localization and Temporal Feature Aggregation197
A novel method integrated with somatic topological line localization (TLL) and temporal feature aggregation for detecting multi-scale pedestrians and a post-processing scheme based on Markov Random Field (MRF) is introduced to eliminate ambiguities in occlusion cases.
- A Free Lunch for Unsupervised Domain Adaptive Object Detection without Source Data193
Self-entropy descent (SED) is a metric proposed to search an appropriate confidence threshold for reliable pseudo label generation without using any handcrafted labels in object detection, where completely clean labels are still unattainable.
- Edit Probability for Scene Text Recognition192
A novel method called edit probability (EP) for scene text recognition tries to effectively estimate the probability of generating a string from the output sequence of probability distribution conditioned on the input image, while considering the possible occurrences of missing/superfluous characters.
- 191
- LoRAMoE: Alleviating World Knowledge Forgetting in Large Language Models via MoE-Style Plugin186
Experimental results show that, as the instruction data increases, Lo-RAMoE can significantly improve the ability to process downstream tasks, while maintaining the world knowledge stored in the LLM.
- Spatio-Temporal Deformable Convolution for Compressed Video Quality Enhancement180
This paper proposes a fast yet effective method for compressed video quality enhancement by incorporating a novel Spatio-Temporal Deformable Fusion (STDF) scheme to aggregate temporal information and achieves the state-of-the-art performance of compressed videoquality enhancement in terms of both accuracy and efficiency.
- Counterfactual Critic Multi-Agent Training for Scene Graph Generation170
CMAT is a multi-agent policy gradient method that frames objects into cooperative agents, and then directly maximizes a graph-level metric as the reward, and uses a counterfactual baseline that disentangles the agent-specific reward by fixing the predictions of other agents.
- Scene Dynamics: Counterfactual Critic Multi-Agent Training for Scene Graph Generation.167
CMAT is a multi-agent policy gradient method that frames objects into cooperative agents, and then directly maximizes a graph-level metric as the reward, and uses a counterfactual baseline that disentangles the agent-specific reward by fixing the predictions of other agents.
- Learning Domain Adaptive Object Detection with Probabilistic Teacher144
A simple yet effective framework, termed as Probabilistic Teacher (PT), which aims to capture the uncertainty of unlabeled target data from a gradually evolving teacher and guides the learning of a student in a mutually beneficial manner, and proposes to leverage the uncertainty-guided consistency training to promote classification adaptation and localization adaptation.
- Can Abnormality be Detected by Graph Neural Networks?137
A novel Adaptive Multi-frequency Graph Neural Network (AMNet), aiming to capture both low-frequency and high-frequency signals, and adaptively combine signals of different frequencies is proposed, comparing with several state-of-the-art baseline methods.
- Text Perceptron: Towards End-to-End Arbitrary-Shaped Text Spotting132
This paper proposes an end-to-end trainable text spotting approach named Text Perceptron, which unites text detection and the following recognition part into a whole framework, and helps the whole network achieve global optimization.
- 119
- Self-Domain Adaptation for Face Anti-Spoofing119
A self-domain adaptation framework to leverage the unlabeled test domain data at inference and a meta-learning based adaptor learning algorithm is proposed using the data of multiple source domains at the training step to learn a better adaptor.
- TRIE: End-to-End Text Reading and Information Extraction for Document Understanding118
This paper proposes a unified end-to-end text reading and information extraction network, where the two tasks can reinforce each other and the multimodal visual and textual features of text reading are fused for information extraction and in turn, the semantics in information extraction contribute to the optimization of text read.
- Collaborative Spatiotemporal Feature Learning for Video Action Recognition112
A novel neural operation is proposed which encodes spatiotemporal features collaboratively by imposing a weight-sharing constraint on the learnable parameters by performing 2D convolution along three orthogonal views of volumetric video data, which learns spatial appearance and temporal motion cues respectively.
- LGPMA: Complicated Table Structure Recognition with Local and Global Pyramid Mask Alignment104
This paper proposes the framework of Local and Global Pyramid Mask Alignment, which adopts the soft pyramid mask learning mechanism in both the local and global feature maps and allows the predicted boundaries of bounding boxes to break through the limitation of original proposals.
- SAViT: Structure-Aware Vision Transformer Pruning Via Collaborative Optimization98
Joint importance is introduced, which integrates essential structural-aware interactions between components for the first time, to perform collaborative pruning in ViTs, and outperforms the existing state-of-the-art approaches on ImageNet.
- TransForensics: Image Forgery Localization with Dense Self-Attention98
This work introduces TransForensics, a novel image forgery localization method inspired by Transformers that not only can capture discriminative representations and obtain high-quality mask predictions but is also not limited by tampering types and patch sequence orders.
- Label Matching Semi-Supervised Object Detection97
A novel label assignment mechanism for self-training framework, namely proposal self-assignment, which injects the proposals from student into teacher and generates accurate pseudo labels to match each proposal in the student model accordingly is presented.
- MANGO: A Mask Attention Guided One-Stage Scene Text Spotter97
This paper proposes a novel Mask AttentioN Guided One-stage text spotting framework named MANGO, in which character sequences can be directly recognized without RoI operation, and achieves competitive and even new state-of-the-art performance on both regular and irregular text spotting benchmarks.
- Video Relation Detection with Spatio-Temporal Graph93
By combining the model (VRD-GCN) and the proposed association method, the framework for video relation detection achieves the best performance in the latest benchmarks and a series of ablation studies demonstrate the method's effectiveness.
- 90
- All You Need Is a Few Shifts: Designing Efficient Convolutional Neural Networks for Image Classification89
To make this idea feasible, a new and novel basic component named Sparse Shift Layer (SSL) is introduced in this paper and an improved network architecture is redesign to Fully Exploit the limited capacity of neural Network (FE-Net).
- Segregated Temporal Assembly Recurrent Networks for Weakly Supervised Multiple Action Detection88
This paper proposes a segregated temporal assembly recurrent (STAR) network for weakly-supervised multiple action detection and designs a score term called segregated temporal gradient-weighted class activation mapping (ST-GradCAM) fused with attention weights.
- Extreme Network Compression via Filter Group Approximation80
A novel decomposition method based on filter group approximation, which can significantly reduce the redundancy of deep convolutional neural networks (CNNs) while maintaining the majority of feature representation, is proposed.
- Entity Concept-enhanced Few-shot Relation Extraction79
A novel entity CONCEPT-enhanced FEw-shot Relation Extraction scheme (ConceptFERE), which introduces the inherent concepts of entities to provide clues for relation prediction and boost the relations classification performance.
- VSR: A Unified Framework for Document Layout Analysis Combining Vision, Semantics and Relations77
A unified framework VSR is proposed for document layout analysis, combining vision, semantics and relations, which supports both NLP-based and CV-based methods and outperforms previous models by large margins.
- Self-Supervised Noisy Label Learning for Source-Free Unsupervised Domain Adaptation73
A simple yet effective Self-Supervised Noisy Label Learning method, which injects self-supervised learning to impose the intrinsic data structure and facilitate label-denoising in the source domain pre-trained model.
- Cross-Relation Cross-Bag Attention for Distantly-Supervised Relation Extraction73
This paper proposes to conduct multi-instance learning with a novel Cross-relation Cross-bag Selective Attention (C2SA), which leads to noise-robust training for distant supervised relation extractor and employs the sentence-level selective attention to reduce the effect of noisy or mismatched sentences.
- Multi-view Adversarial Discriminator: Mine the Non-causal Factors for Object Detection in Unseen Domains71
A Multi-view Adversarial Discriminator (MAD) based domain generalization model, consisting of a Spurious Correlations Generator (SCG) that increases the diversity of source domain by random augmentation and a Multi-View Domain Classifier (MVDC) that maps features to multiple latent spaces, such that the non-causal factors are removed and the domain-invariant features are purified.
- Adversarial Seeded Sequence Growing for Weakly-Supervised Temporal Action Localization67
A novel weakly-supervised framework by adversarial learning of two modules for eliminating such demerits is proposed, which is designed as a well-designed Seeded Sequence Growing (SSG) Network for progressively extending seed regions to their expected boundaries.
- 66
- 63
- CAKE: A Scalable Commonsense-Aware Framework For Multi-View Knowledge Graph Completion61
A novel and scalable Commonsense-Aware Knowledge Embedding framework to automatically extract commonsense from factual triples with entity concepts and generates commonsense augments effective self-supervision to facilitate both high-quality negative sampling and joint commonsense and fact-view link prediction.
- 57
- 57
- Cascade region proposal and global context for deep object detection56
This paper proposes a novel lightweight cascade structure which can effectively improve RPN proposal quality and improves Faster R-CNN baseline by a large margin and re-implement global context modeling with a few modications.
- 55
- Smart data driven traffic sign detection method based on adaptive color threshold and shape symmetry55
A new traffic sign detection method based on adaptive color threshold segmentation and the hypothesis testing of shape symmetry by leveraging traffic signs and image data is proposed, which has higher detection accuracy and time efficiency than other methods, and better robustness under complex traffic environment.
- Unimodal-Concentrated Loss: Fully Adaptive Label Distribution Learning for Ordinal Regression50
This paper proposes a novel loss function for fully adaptive label distribution learning, namely unimodal-concentrated loss, which is integrated into the proposed concentrated loss to make the predicted distribution maximize at the ground-truth and vary according to the predicting uncertainty.
- 50
- 50
- Multi-Turn Video Question Answering via Multi-Stream Hierarchical Attention Context Network50
This paper proposes the hierarchical attention context network for context-aware question understanding by modeling the hierarchically sequential conversation context structure and develops the multi-stream spatio-temporal attention network for learning the joint representation of the dynamic video contents and context- aware question embedding.
- 50
- Breast mass detection in digital mammography based on anchor-free architecture49
A one-stage object detection architecture, called Breast Mass Detection Network (BMassDNet), based on anchor-free and feature pyramid which makes the detection of breast masses of different sizes well adapted and introduces a truncation normalization method and combines it with adaptive histogram equalization to enhance the contrast between the breast mass and the surrounding environment.
- Learning Incremental Triplet Margin for Person Re-Identification49
The margin between positive and negative pairs of triplets is explored and it is proved that large margin is beneficial and a novel multi-stage training strategy which learns incremental triplet margin and improves triplet loss effectively is proposed.
- 48
- 47
- 47
- A Practical Convolutional Neural Network as Loop Filter for Intra Frame46
A CNN as the loop filter for intra frames is proposed and a single CNN model with low redundancy to adapt to decoded frames with different qualities and ensure consistency is designed.
- InsPose: Instance-Aware Networks for Single-Stage Multi-Person Pose Estimation45
This work proposes an instance-aware module to adaptively adjust (part of) the network parameters for each instance to significantly increase the capacity and adaptive-ability of the network for recognizing various poses, while maintaining a compact end-to-end trainable pipeline.
- 44
- You Only Recognize Once44
This paper proposes a fast and robust video text spotting framework by only recognizing the localized text one-time instead of frame-wisely recognition, and develops a novel text recommender for selecting the highest-quality text from text streams and only recognizes the selected ones.
- 44
- 42
- 42
- 40
- 39
- 38
- 38
- FREE: A Fast and Robust End-to-End Video Text Spotter38
A fast and robust end-to-end video text spotting framework named FREE is proposed by only recognizing the localized text stream one-time instead of frame-wise recognition, which greatly speeds up the text spotting process, and also achieves the remarkable state-of-the-art.
- MAFF-Net: Filter False Positive for 3D Vehicle Detection with Multi-modal Adaptive Feature Fusion37
Two end-to-end trainable single-stage multi-modal feature fusion approaches to adaptively combine RGB and point cloud modalities are proposed and can provide competitive results compared to the published state-of-the-art multi- modal methods in the KITTI benchmark.
- HERO: HiErarchical spatio-tempoRal reasOning with Contrastive Action Correspondence for End-to-End Video Object Grounding35
This paper introduces the weakly-supervised contrastive learning that classifies the video as action-consistent and action-independent frames relying on the video-caption action semantic correspondence and designs the hierarchical reasoning layers to decouple fully connected multi-head attention and remove the redundant interfering correlations.
- SPIN: Structure-Preserving Inner Offset Network for Scene Text Recognition34
A new learnable geometric-unrelated rectification, Structure-Preserving Inner Offset Network (SPIN), which allows the color manipulation of source data within the network, giving neural networks the ability to actively transform input intensity rather than only the spatial rectification.
- AutoETER: Automated Entity Type Representation for Knowledge Graph Embedding34
This work develops a novel KGE framework with Automated Entity TypE Representation (AutoETER), which learns the latent type embedding of each entity by regarding each relation as a translation operation between the types of two entities with a relation-aware projection mechanism.
- Contrastive Learning-Based Embedder for the Representation of Tandem Mass Spectra33
A novel contrastive learning-based method for the representation of MS/MS spectra, called CLERMS, which is based on transformer architecture equipped with a sinusoidal embedder and a novel loss function composed of InfoNCE loss and MSE loss has been proposed for the attainment of good embedding from the peak information and the metadata.
- Multichannel Attention Refinement for Video Question Answering33
Appearance, motion, and audio features are extracted from the video, and question-guided attentions are refined to generate the expressive clues that support the correct answer in VideoQA.
- 33
- Estimation of crowd density in surveillance scenes based on deep convolutional neural network31
A new crowddensity estimation method by deep convolutional neural network (ConvNet) is proposed and a new dataset including 31 crowd Subway-carriage scenes with over 160K density annotated images is introduced to better evaluate the accuracy of cross-scene crowd density estimation methods.
- Efficient distributed reachability querying of massive temporal graphs30
An efficient index, called Temporal Vertex Labeling (TVL), is proposed, which is a labeling scheme for distributed temporal graphs, and algorithms that exploit TVL to achieve efficient support for distributed reachability querying over temporal graphs in Pregel-like systems are presented.
- Rethinking the Approximation Error in 3D Surface Fitting for Point Cloud Normal Estimation29
Two basic design principles are presented, which applies the Z-direction Transform to rotate local patches for a better surface fitting with a lower approximation error and models the error of the normal estimation as a learnable term, and are implemented using deep neural networks.
- 29
- 27
- 26
- 2C-Net: integrate image compression and classification via deep neural network24
A versatile framework is studied to explore the common feature representations for both image compression and classification and shows that the proposed method outperforms the conventional codecs like BPG and JPEG2000 in compression efficiency, while achieving acceptable accuracy on different image classification datasets without image reconstruction.
- 23
- Universal Domain Adaptive Object Detector23
US-DAF is proposed, namely Universal Scale-Aware Domain Adaptive Faster RCNN with Multi-Label Learning, to reduce the negative transfer effect during training while maximizing transferability as well as discriminability in both domains under a variety of scales.
- Towards Comprehensive Representation Enhancement in Semantics-Guided Self-supervised Monocular Depth Estimation22
This work proposes an attention-based module to enhance task-specific feature by addressing their feature uniqueness within instances and proposes a metric learning based approach to accomplish comprehensive enhancement on depth feature by creating a separation between instances in feature space.
- Distributed time-respecting flow graph pattern matching on temporal graphs22
A new notion so-called time-respecting flow graph is proposed, in which all paths spreading from one vertex to another are time-respecting (i.e., a series of contacts with non-decreasing time), and one vertex is distinguished as the root from which other vertices can be reached via a time- respecting path.
- 21
- AutoETER: Automated Entity Type Representation with Relation-Aware Attention for Knowledge Graph Embedding.21
This work develops a novel KGE framework with Automated Entity TypE Representation (AutoETER), which learns the latent type embedding of each entity by regarding each relation as a translation operation between the types of two entities with a relation-aware projection mechanism.
- A Lightweight Siamese Neural Network for Building Change Detection Using Remote Sensing Images20
LightCDNet, a lightweight Siamese neural network for BuCD, is introduced and it is demonstrated that LightCDNet outperforms several state-of-the-art methods in accuracy and efficiency.
- 20
- PROSE: Graph Structure Learning via Progressive Strategy19
It is argued that current graph structure learning methods still pay no regard to the status of nodes and just judge all of their connections simultaneously using a monotonous standard, which will lead to indeterminacy and instability in the optimization process.
- Learning Multiple Explainable and Generalizable Cues for Face Anti-Spoofing19
Four mainly used cues by humans are introduced as auxiliary supervision including the boundary of spoof medium, moiré pattern, reflection artifacts and facial depth in addition to the binary supervision in a novel framework to learn multiple explainable and generalizable cues (MEGC) for face anti-spoofing.
- Slimmable Domain Adaptation19
A simple framework to improve cross-domain generalization with a weight-sharing model bank, from which models of different capacities can be sampled to accommodate different accuracy-efficiency trade-offs, and augment the existing bi-classifier domain confusion architecture into an Optimization-Separated Tri-Classifier counterpart.
- Weakly Supervised Regional and Temporal Learning for Facial Action Unit Recognition19
This work proposes an end-to-end trainable framework named weakly supervised regional and temporal learning (WSRTL) for AU recognition and proposes two auxiliary AU related tasks to bridge the gap between limited annotations and the model performance in a self-supervised manner via the unlabeled data.
- 19
- Arbitrary-Scale Point Cloud Upsampling by Voxel-Based Network with Latent Geometric-Consistent Learning18
A density-guided grid resampling method is developed to generate high-fidelity points while effectively avoiding sampling outliers and an auxiliary training supervision is presented to enforce the latent geometric consistency among local surface patches.
- 18
- Video question answering via grounded cross-attention network learning18
A novel hierarchical cross-attention mechanism of mutual attention learning for video question answering, named as GCANet is proposed, which helps to learn the joint question-video representation based on both rough representation and grounded representation of video for video questioning.
- Multi-modal Action Chain Abductive Reasoning17
The proposed model significantly outperforms existing video-language models in terms of effectiveness on the newly created MAR dataset and design a novel symbolic module that can complete strict derivation of the progressive action chain layer by layer.
- TRIE++: Towards End-to-End Information Extraction from Visually Rich Documents17
A unified end-to-end information extraction framework from visually rich documents, where text reading and information extraction can reinforce each other via a well-designed multi-modal context block is proposed.
- 17
- 17
- 17
- 17
- Reciprocal Feature Learning via Explicit and Implicit Tasks in Scene Text Recognition15
This work excavate the implicit task, character counting within the traditional text recognition, without additional labor annotation cost, and designs a two-branch reciprocal feature learning framework in order to adequately utilize the features from both the tasks.
- 15
- Multi-Tasking U-Shaped Network for Benign and Malignant Classification of Breast Masses15
A Multi-Tasking U-shaped Network (MT-UNet), which contains three key ideas that can well adapt to the heterogeneity of breast masses, and experimental results show that the proposed method can obtain a competitive performance.
- 15
- Unsupervised Image Classification for Deep Representation Learning15
This work proposes an unsupervised image classification framework without using embedding clustering, which is very similar to standard supervised training manner, and aims to make this framework more simple and elegant without performance decline.
- 14
- 14
- Unsupervised object detection with scene-adaptive concept learning14
This work proposes a novel scene-adaptive evolution unsupervised video object detection algorithm that can decrease the impact of scene changes through the concept of object groups, and builds the visual knowledge dictionary of object concepts by clustering the proposals.
- A Layer Decomposition-Recomposition Framework for Neuron Pruning towards Accurate Lightweight Networks14
A novel Layer Decomposition-Recomposition Framework (LDRF) for neuron pruning is proposed, by which each layer's output information is recovered in an embedding space and then propagated to reconstruct the following pruned layers with useful information preserved.
- Arbitrarily-Oriented Text Recognition.14
The arbitrary orientation network (AON) is developed to capture the deep features of irregular texts (e.g. arbitrarily-oriented, perspective or curved), which are combined into an attention-based decoder to generate character sequence.
- Visible and near‐infrared image fusion based on information complementarity13
The experimental results demonstrate that the proposed IFC approach outperforms the state‐of‐the‐art in both avoiding artificial colours as well as effectively utilising information complementarity.
- MTTPRE13
A novel Multi-scale spatial-temporal model for Travel Time Prediction, abbreviated as MTTPRE, which outperforms all the competitors with significant improvement and remarkable robustness.
- 13
- 3D CNN-based fingerprint anti-spoofing through optical coherence tomography12
An OCT fingerprint anti-spoofing method based on a 3D convolutional neural network (CNN) is proposed, considering the spatial continuity of 3D biometrics in fingertips, and achieves the best results compared with classic networks.
- Self-Supervised Regional and Temporal Auxiliary Tasks for Facial Action Unit Recognition12
This work proposes two auxiliary AU related tasks to bridge the gap between limited annotations and the model performance in a self-supervised manner via the unlabeled data and designs a task of RoI inpainting to recover the randomly cropped AU patches to enhance the discrimination of regional features with AU relation embedding.
- IROS 2019 Lifelong Robotic Vision: Object Recognition Challenge [Competitions]12
This report summarizes the IEEE/ RSJ International Conference on Intelli gent Robots and Systems (IROS) 2019 Lifelong Robotic Vision Competition (Lifelong Object Recognition Chal lenge) with the data set, rules, methods, and results from the top eight finalists (of over 150 teams) (Figure 2).
- Enhancing Multi-modal Features Using Local Self-attention for 3D Object Detection11
EMMF-Det uses self-attention mechanism to do feature re-weighting on these two modalities interactively, which can enchance the features with color, texture and localiztion information provided by Li-DAR and camera signals.
- 11
- 11
- Learned Quality Enhancement via Multi-Frame Priors for HEVC Compliant Low-Delay Applications11
A Quality Enhancement Network (QENet) is developed to reduce the video compression artifacts, leveraging the spatial and temporal priors generated by respective multi-scale convolutions spatially and warped temporal predictions in a recurrent fashion temporally.
- Unified Normalization for Accelerating and Stabilizing Transformers10
Unified Normalization is proposed, which can speed up the inference by being fused with other linear operations and achieve comparable performance on par with LN, and strives to boost performance by calibrating the activation and gradient statistics with a tailored fluctuation smoothing strategy.
- End-to-End Compound Table Understanding with Multi-Modal Modeling10
A framework named CTUNet is proposed to integrate the compromised visual, semantic, and position features with a graph attention network, which can solve the table recognition task and the challenging table understanding task as a whole.
- 10
- Demodalizing Face Recognition with Synthetic Samples10
This paper proposes a demodalizing face recognition training architecture for the first time and provides a feasible method for recognition training using synthetic samples, and validates the effectiveness of the approach on various benchmarks of large-scale face recognition and outperform the previous methods.
- Efficient Video Scene Text Spotting: Unifying Detection, Tracking, and Recognition.10
To efficiently spot video text, the method localizes and tracks text in each frame, and recognizes each tracked text stream one-time with a text region quality scoring mechanism instead of identifying the cropped text regions one-by-one.
- 10
- Unsupervised Domain Adaptation for Video Object Grounding with Cascaded Debiasing Learning9
This paper addresses the Unsupervised Domain Adaptation (UDA) for the dense frame prediction task - Video Object Grounding and measures Grounding Coordinate Offset based on the semantic distance of the model's prediction across domains, based on which it alleviate source-domain bias in the target model through adversarial learning.
- 9
- Hierarchical Temporal Fusion of Multi-grained Attention Features for Video Question Answering9
This work proposes the multi-granularity temporal attention network that enables to search for the specific frames in a video that are holistically and locally related to the answer.
- 8
- Memory Graph with Message Rehearsal for Multi-Turn Dialogue Generation8
A Memory Graph with Message Rehearsal (MGMR) is proposed for dialogue generation based on the cognitive process of human memory and innovatively design a memory graph in long-term memory to construct the relations among words, utterances and knowledge.
- 8
- A Genetic-Algorithm-Based Sparse Planar Array Optimization Method for 77 GHz High-Resolution Imaging8
- Accelerating Dynamic Network Embedding with Billions of Parameter Updates to Milliseconds7
The Dynamic Adjacency Matrix Factorization (DAMF) algorithm is proposed, which achieves an efficient and accurate dynamic network embedding by rotating and scaling the coordinate system where the network embeding resides with no more than the number of edge modifications changes of node embeddings.
- 7
- 7
- Towards Robust Model Reuse in the Presence of Latent Domains7
The MRL (Model Reuse for multiple Latent domains) method, where both domain characteristics and pre-trained models are considered for the exploration of instances in the target task and the overall considerations are packed in a bi-level optimization framework with a reliable generalization.
- Neural Inheritance Relation Guided One-Shot Layer Assignment Search7
A neural inheritance relation among the networks with different layer assignments is discovered, that is, the optimal layer assignments for deeper networks always inherit from those for shallow networks, and an efficient one-shot layer assignment search approach via inherited sampling is proposed.
- Insert or Attach: Taxonomy Completion via Box Embedding6
The framework, TaxBox, leverages box containment and center closeness to design two specialized geometric scorers within the box embedding space, tailored for insertion and attachment operations and can effectively capture intrinsic relationships between concepts by optimizing on a granular box constraint loss.
- Fast Updating Truncated SVD for Representation Learning with Sparse Matrices6
This work introduces a novel method for dynamically approximating the truncated SVD of a sparse and temporally evolving matrix that leverages sparsity in the orthogonalization process of augmented matrices and utilizes an extended decomposition to independently store projections in the column space of singular vectors.
- Edge-Preserving Texture-Based Semantic Segmentation for Ultrahigh Resolution Images in Agricultural Scene6
A texture-based semantic segmentation method for ultrahigh-resolution images in agricultural scenes, which incorporates an encoding layer to capture textural information and exhibits strong advantages in noise suppression and edge preservation.
- 6
- 6
- 6
- 6
- Transductive Clip with Class-Conditional Contrastive Learning6
This work proposes Transductive CLIP, a novel framework for learning a classification network with noisy labels from scratch, and proposes a class-conditional contrastive learning mechanism to mitigate the reliance on pseudo labels and boost the tolerance to noisy labels.
- 6
- 6
- Photo Stream Question Answer6
This paper presents a new visual question answering (VQA) task -- Photo Stream QA, which aims to answer the open-ended questions about a narrative photo stream, and proposes an end-to-end baseline (E-TAA), which provides promising results outperforming all the other baseline methods.
- 6
- 6
- 6
- Learning Expressive And Generalizable Motion Features For Face Forgery Detection5
An effective sequence-based forgery detection framework based on an existing video classification method is proposed, and an alternative motion consistency block instead of the original motion features module is proposed to make the learned features more generalizable.
- 5
- 5
- Multi-Level Adaptive Region of Interest and Graph Learning for Facial Action Unit Recognition5
A novel multi-level adaptive ROI and graph learning (MARGL) framework to tackle the problem of limited representation capacity of regional features in facial action unit (AU) recognition tasks and significantly outperforms the previous state-of-the-art methods.
- 5
- 5
- 5
- 5
- Robust fuzzy tracking control for flexible air-breathing hypersonic vehicles with measurement noises5
A robust fuzzy tracking control scheme for the longitudinal dynamics of the hypersonic vehicle is constructed by utilizing type-2 fuzzy adaptive technique and dynamic surface control (DSC) approach.
- An End-to-End Audio Classification System Based on Raw Waveforms and Mix-Training Strategy5
An end-to-end audio classification system based on raw waveforms and a mix-training strategy to break the performance limitation caused by the amount of training data and exceeds the state-of-the-art multi-level attention model.
- A Solution to Co-occurrence Bias in Pedestrian Attribute Recognition: Theory, Algorithms, and Improvements4
A novel concept of attributes-disentangled feature learning is formulated, by which the mutual information among features of different attributes is minimized, ensuring the recognition of an attribute independent to the presence of others.
- 4
- Single Domain Dynamic Generalization for Iris Presentation Attack Detection4
A Single Domain Dynamic Generalization (SDDG) framework, which simultaneously exploits domain-invariant and domain-specific features on a per-sample basis and learns to generalize to various unseen domains with numerous natural images is proposed.
- 4
- Self-distilled Knowledge Delegator for Exemplar-free Class Incremental Learning4
This paper introduces a so-called knowledge delegator, which is capable of transferring knowledge from the trained model to a randomly re-initialized new model by generating informative samples, and achieves comparable performance to some exemplar-based methods without accessing any exemplars.
- 4
- 4
- Learning embeddings of a heterogeneous behavior network for potential behavior prediction4
A novel network embedding method is proposed, human-action-attribute-aware heterogeneous network embedding (a4HNE), which jointly considers structural proximity, attribute resemblance, and heterogeneity fusion and outperforms other similar methods on various heterogeneous information network mining tasks for potential behavior prediction.
- 4
- 3
- 3
- Liberate Pseudo Labels from Over-Dependence: Label Information Migration on Sparsely Labeled Graphs3
This paper proposes a novel Graph Label Information Migration framework (GLIM) that uses abundant training process information to find more reliable node labels and improve the model robustness against label noise and proposes a label migration module (LMM) that migrates label information from label-dense regions to label-sparse regions by a spectral based graph matching algorithm.
- 3
- 3
- 3
- 3
- 3
- Associative affinity network learning for multi-object tracking3
A simple multi-object tracker that achieves competitive performance on the public MOT16 and MOT17 test datasets is proposed, and can also perform single-object tracking.
- 3
- A Strong Baseline for Semi-Supervised Incremental Few-Shot Learning3
A novel paradigm containing two parts is proposed: a well-designed meta-training algorithm for mitigating ambiguity between base and novel classes caused by unreliable pseudo labels and a model adaptation mechanism to learn discriminative features for novel classes while preserving base knowledge using few labeled and all the unlabeled data.
- Text Recognition in Real Scenarios with a Few Labeled Samples3
A few-shot adversarial sequence domain adaptation (FASDA) approach to build sequence adaptation between the synthetic source domain and a specific target domain (with only some or a few real labeled samples) and achieves the sequence-level adaptation with even a small number of labeled samples in the target domain.
- 3
- 3
- 3
- Efficient Similarity Search on Quasi-Metric Graphs3
A new notion called quasi-metric graph is introduced that connects metric data using a graph, and two simple efficient approaches are proposed, which traverse the quasi-Metric graph following the best-first and the breadth-first paradigms, respectively, and utilize the triangle inequality to prune unnecessary evaluation.
- 3
- 2
- 1st Place Solution for ECCV 2022 OOD-CV Challenge Object Detection Track2
A simple yet effective Generalize-then-Adapt (G&A) framework, which is composed of a two-stage domain generalization part and a one-stagedomain adaptation part, to solve the problem in object detection track of the OOD-CV challenge.
- Cognitive-inspired Graph Redundancy Networks for Multi-source Information Fusion2
Inspired by Multisensory Processing theory in cognitive neuroscience, Graph Redundancy Network (GRN) is proposed that learns a suitable representation space that maximizes multi-source interactions and encodes the redundant and complementary information according to Graph Intersection and Difference of their graph structures.
- 2
- E2-AEN: End-to-End Incremental Learning with Adaptively Expandable Network2
This paper proposes an end-to-end trainable adaptively expandable network named E2-AEN, which dynamically generates lightweight structures for new tasks without any accuracy drop in previous tasks, and introduces a novel sparsity-activation regularization to encourage the model to learn discriminative features with limited parameters.
- 2
- 2
- 2
- HRI Team @ TRECVID 2016: Surveillance event detection.2
Two neural networks are utilized: Region Proposal Network and Detection Neural Network are trained end-to-end to generate high-quality region proposals, then Detection Neural Networks is used to determine whether the region proposals are target or not.
- 1
- GraphFADE: Field-aware Decorrelation Neural Network for Graphs with Tabular Features1
This work proposes Field-aware Decorrelation Neural Network for graphs with tabular features (GraphFADE), a novel framework that directly optimizes the overcorrelation problem for graphs that are heterogeneous with multiple feature fields.
- 1st Place Solution for ECCV 2022 OOD-CV Challenge Image Classification Track1
This work proposes a simple yet effective Mask-Level Copy-Paste data augmentation strategy to enhance out-of-distribution generalization ability so as to resist shape, pose, context, texture, occlusion, and weather domain shifts in this challenge.
- 1
- 1
- Semi-Supervised Ranking for Object Image Blur Assessment1
This work proposes to label the rank relationships between pairwise images rather their quality levels, and establish a large-scale realistic face image blur assessment dataset with reliable labels, and proposes a self-supervised method based on quadruplet ranking consistency to leverage the unlabeled data more effectively.
- 1
- 1
- 2nd Place Solution for ICCV 2021 VIPriors Image Classification Challenge: An Attract-and-Repulse Learning Approach1
A new framework, termed Attract-and-Repulse, which consists of Contrastive Regularization to enrich the feature representations, Symmetric Cross Entropy to balance the fitting for different classes and Mean Teacher to calibrate label information is proposed and achieves the second place in ICCV 2021 VIPriors Image Classification Challenge.
- 1
- UWC: Unit-wise Calibration Towards Rapid Network Compression1
This work defines several adjacent layers as a Basic-Unit, and presents a unit-wise post-training algorithm which can minimize quantization error, and achieves near-original accuracy on ImageNet and COCO when quantizing FP32 models to INT4 and INT3.
- 1
- Modulating Localization and Classification for Harmonized Object Detection1
This work proposes a mutual learning framework to modulate the two tasks of localization and classification, which achieves a significant performance gain over the baseline detectors on the COCO dataset.
- 1
- 1
- 1
- –
- –
- –
- Towards Deployment-Efficient and Collision-Free Multi-Agent Path Finding (Student Abstract)–
The merits of planning-based and RL-based MAPF methods are combined to propose a deployment-efficient and collision-free MAPF algorithm that can be deployed efficiently and prevent collisions entirely.
- –
- –
- Cognize Yourself–
This is the first study to combine graph pre-training with cognitive theory to build a cognition-aware model and CORE consists of cognizing network and differentiating process, the former cognizes a core graph which stands for the essential structure of the graph, and the latter allows it to differentiate into several task-specific graphs for different tasks.
- –
- Robust Vectorized Surface Reconstruction with 2D-3D Joint Optimization–
A robust pipeline to create vectorized models from LiDAR point clouds without the assumption of watertight polygonal surfaces is proposed and shows that the proposed method obtains compact models with high fidelity and efficiency on either precise or defect-laden data compared with other state-of-the-art approaches.
- –
- –
- KRNet: Towards Efficient Knowledge Replay–
A novel and efficient knowledge recording network (KRNet) which directly maps an arbitrary sample identity number to the corresponding datum and requires significantly (400×) less storage cost for the latent codes and can be trained without the encoder sub-network.
- –
- –
- –
- –
- –
- Posterior-regularized REINFORCE for Instance Selection in Distant Supervision–
This paper adopts posterior regularization (PR) to integrate some domain-specific rules in instance selection using REINFORCE and remarkably improves the performance of the relation classifier trained on cleaned distant supervision dataset as well as the efficiency of the REIN FORCE training.
- –
- –
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 2026-10-12. 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.