Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks162 from public data
- 653
- 422
- Learning Semantic Concepts and Order for Image and Sentence Matching342
A semantic-enhanced image and sentence matching model is proposed, which can improve the image representation by learning semantic concepts and then organizing them in a correct semantic order.
- ETPNav: Evolving Topological Planning for Vision-Language Navigation in Continuous Environments292
A new navigation framework, ETPNav, which focuses on two critical skills: the capability to abstract environments and generate long-range navigation plans, and the ability of obstacle-avoiding control in continuous environments.
- Bidirectional recurrent convolutional networks for multi-frame super-resolution273
This work proposes a bidirectional recurrent convolutional network for efficient multi-frame SR, different from vanilla RNNs, which has a low computational complexity and runs orders of magnitude faster than other multi- frame methods.
- Box-Driven Class-Wise Region Masking and Filling Rate Guided Loss for Weakly Supervised Semantic Segmentation257
This paper introduces a box-driven class-wise masking model (BCM) to remove irrelevant regions of each class and proposes a filling rate guided adaptive loss (FR-Loss) to help the model ignore the wrongly labeled pixels in proposals.
- Improving Description-Based Person Re-Identification by Multi-Granularity Image-Text Alignments238
A Multi-granularity Image-text Alignments (MIA) model is proposed to alleviate the cross-modal fine-grained problem for better similarity evaluation in description-based person Re-id and obtains the state-of-the-art performance on the CUHK-PEDES dataset.
- Instance-Aware Image and Sentence Matching with Selective Multimodal LSTM235
Extensive experiments show that the proposed selective multimodal Long Short-Term Memory network can well match image and sentence with complex content, and achieve the state-of-the-art results on two public benchmark datasets.
- 231
- 209
- Rethinking the Heatmap Regression for Bottom-up Human Pose Estimation205
The scale-adaptive heatmap regression (SAHR) method, which can adaptively adjust the standard deviation for each keypoint, is proposed, which largely improves the accuracy of bottom-up human pose estimation and is comparable with the performances of most top-down methods.
- 198
- M3: Multimodal Memory Modelling for Video Captioning159
A Multimodal Memory Model (M3) is proposed to describe videos, which builds a visual and textual shared memory to model the long-term visual-textual dependency and further guide visual attention on described visual targets to solve visual- Textual alignments.
- 147
- GaitNet: An end-to-end network for gait based human identification146
This paper integrates an end-to-end network for gait recognition, named GaitNet, composed of two convolutional neural networks: one corresponds to gait segmentation, and the other corresponds to classification, which greatly simplifies the traditional step-by-step manner and is thus much more efficient for practical applications.
- Cross-Modal Ranking with Soft Consistency and Noisy Labels for Robust RGB-T Tracking145
A novel approach to suppress background effects for RGB-T tracking by integrating the soft cross-modality consistency into the ranking model which allows the sparse inconsistency to account for the different properties between these two modalities.
- 143
- 143
- Part-Level Graph Convolutional Network for Skeleton-Based Action Recognition120
A novel Part-Level Graph Convolutional Network (PL-GCN) is proposed to capture part-level information of skeletons, which aims at learning high-level relations between body parts while the PA block aims at highlighting the important body parts in the action.
- Neighbor-view Enhanced Model for Vision and Language Navigation99
This work proposes a multi-module Neighbor-View Enhanced Model (NvEM) to adaptively incorporate visual contexts from neighbor views for better textual-visual matching of existing methods.
- Target tracking based on a distributed particle filter in underwater sensor networks85
Based on a distributed particle filter, two tracking algorithms are proposed for tracking mobile targets in cluster-based underwater sensor networks (USNs) and results indicate that one tracking algorithm achieves higher tracking accuracy while the other achieves dramatic reduction of communication cost, energy cost, and tracking response time.
- Advances in Online Audio-Visual Meeting Transcription84
A system that generates speaker-annotated transcripts of meetings by using a microphone array and a 360-degree camera and an online audio-visual speaker diarization method that leverages face tracking and identification, sound source localization, speaker identification and, if available, prior speaker information for robustness to various real world challenges is described.
- Regularized Graph Structure Learning with Semantic Knowledge for Multi-variates Time-Series Forecasting83
This paper proposes Regularized Graph Structure Learning (RGSL) model to incorporate both explicit prior structure and implicit structure together, and learn the forecasting deep networks along with the graph structure.
- Constraint-Aware Zero-Shot Vision-Language Navigation in Continuous Environments77
A Constraint-Aware Navigator (CA-Nav), which reframes zero-shot VLN-CE as a sequential, constraint-aware sub-instruction completion process, and achieves the state-of-the-art performance on two VLN-CE benchmarks, surpassing the previous best method by 12% and 13% in Success Rate on the validation unseen splits of R2R-CE and RxR-CE, respectively.
- Relational Prototypical Network for Weakly Supervised Temporal Action Localization73
This paper proposes a weakly supervised temporal action localization method on untrimmed videos based on prototypical networks that outperforms the state-of-the-art methods on action-background separation and action relation construction.
- 71
- Cyclic Differentiable Architecture Search62
This work proposes new joint optimization objectives and a novel Cyclic Differentiable ARchiTecture Search framework, dubbed CDARTS, which enables the evolution of the architecture to fit the final evaluation network.
- 61
- A hierarchical contextual attention-based network for sequential recommendation59
A Hierarchical Contextual Attention-based network is proposed that can relieve the limitation of ordered modeling of RNN, model the complicated correlations among recent factors, and strengthen the short-term interest.
- Efficient Token-Guided Image-Text Retrieval With Consistent Multimodal Contrastive Training58
A Token-Guided Dual Transformer (TGDT) architecture which consists of two homogeneous branches for image and text modalities, respectively, is proposed for image-text retrieval and achieves state-of-the-art retrieval performances with extremely low inference time when compared with representative recent approaches.
- GR-MG: Leveraging Partially-Annotated Data via Multi-Modal Goal-Conditioned Policy54
A novel progress-guided goal image generation model which injects task progress information into the generation process and enables GR-MG to leverage large amounts of partially-annotated data while still using languages to flexibly specify tasks.
- Learning view invariant gait features with Two-Stream GAN53
A Two-Stream Generative Adversarial Network (TS-GAN) for cross-view gait recognition that combines the two streams to learn final identities and adds a pixel-wise loss along with the generators of GAN to restore the gait details in pixel-level.
- Skeleton-Based Relational Modeling for Action Recognition.50
An Attentional Recurrent Relational Network-LSTM (ARRN-L STM) to simultaneously model spatial configurations and temporal dynamics in skeletons for action recognition and introduces an adaptive attentional module for focusing on potential discriminative parts of the skeleton towards a certain action.
- 1st Place Solutions for RxR-Habitat Vision-and-Language Navigation Competition (CVPR 2022)48
The winning entry of the RxR-Habitat Competition in CVPR 2022 is presented, which takes several recent advances of Vision-and-Language Navigation to improve the performance such as pretraining based on large-scale synthetic in-domain dataset, environment-level data augmentation and snapshot model ensemble.
- 47
- 43
- Textual Dependency Embedding for Person Search by Language42
This work focuses on the long-distance dependencies in a sentence for better textual encoding, and proposes the Textual Dependency Embedding (TDE) method, which has obtained the state-of-the-art performance on the CUHK-PEDES and Flickr30K datasets.
- End-to-End Alternating Optimization for Real-World Blind Super Resolution40
This work designs two convolutional neural modules, namely Restorer and Estimator, which can estimate the degradation and restore the SR image in a single model and can largely outperform state-of-the-art methods and produce more visually favorable results.
- 39
- Cross-modal Co-occurrence Attributes Alignments for Person Search by Language38
A novel cross-modal matching method named Cross- modal Co-occurrence Attributes Alignments (C2A2), which can better deal with noise and obtain significant improvements in retrieval performance for person search by language, and significantly outperforms the existing solutions.
- 37
- AI in Human-computer Gaming: Techniques, Challenges and Opportunities36
A survey of recent successful game AIs, covering board game AIS, card game A is, first-person shooting game Ais, and real-time strategy game AI, to compare the main difficulties among different kinds of games and the corresponding techniques utilized for achieving professional human-level AIs.
- 35
- 34
- SliceMamba With Neural Architecture Search for Medical Image Segmentation33
SliceMamba features an efficient Bidirectional Slicing and Scanning (BSS) module, which performs bidirectional feature slicing and employs varied scanning mechanisms for sliced features with distinct shapes, and an Adaptive Slicing Search method that automatically identifies the optimal feature slicing method based on the characteristics of the target data.
- 33
- Towards Part-Aware Monocular 3D Human Pose Estimation: An Architecture Search Approach33
This work attempts to build a part-aware 3D pose estimator by searching a set of network architectures so that the model automatically learns to select a suitable architecture to estimate each body part.
- Prediction and Recovery for Adaptive Low-Resolution Person Re-Identification32
This paper proposes a novel Prediction, Recovery and Identification (PRI) model for LR re-id, which adaptively recovers missing details by predicting a preferable scale factor based on the image content, and contains a self-supervised scale factor metric that automatically generates dynamic soft labels.
- Few-Shot Image and Sentence Matching via Gated Visual-Semantic Embedding32
This work focuses on this challenging problem of few-shot image and sentence matching, and proposes a Gated Visual-Semantic Embedding (GVSE) model to deal with it, and achieves the state-of-the-art results on two public benchmark datasets.
- YOLO-CSM-Based Component Defect and Foreign Object Detection in Overhead Transmission Lines31
An interpretability experiment is conducted to reveal the similarity between YOLO-CSM’s attention patterns and that of humans, aiming to explain YOLO-CSM’s advantages in detecting small objects and minor defects in the working environments of power transmission lines.
- Nonlocal spatial attention module for image classification31
A nonlocal spatial attention module (NL-SAM), which collects context information from all pixels to adaptively recalibrate spatial responses in a convolutional feature map and overcomes the limitations of repeating local operations and exports a 2D spatial attention map to emphasize or suppress responses in different locations.
- 29
- Long video question answering: A Matching-guided Attention Model29
A Matching-guided Attention Model (MAM) is proposed which jointly extracts question-related video snippets and predicts answers in a unified framework and uses the attended visual features along with the question to predict the answer in a classification manner.
- 28
- 28
- Causal Intervention for Sparse-View Gait Recognition26
This work forms this overlooked problem as Sparse-View Gait Recognition and provides a comprehensive analysis of it by a Structural Causal Model for causalities among latent features, view distribution, and labels and proposes a simple yet effective method that enables networks to learn a more robust representation among different views.
- 3D Shape Temporal Aggregation for Video-Based Clothing-Change Person Re-identification24
A 3D Shape Temporal Aggregation (3STA) model for video-based clothing-change Re-ID is proposed and an identity-aware 3D shape generation module is introduced that embeds the identity information into the generation of 3D shapes by the joint learning of shape estimation and identity recognition.
- Semi-Supervised Training in Deep Learning Acoustic Model24
It is found that DNN, unfolded RNN, and LSTM-RNN are increasingly more sensitive to labeling errors and the importance sampling has similar impact on all three models with 2~3% relative WER reduction comparing to the random sampling.
- Rapid RNN-T Adaptation Using Personalized Speech Synthesis and Neural Language Generator23
The proposed methodology to a speaker adaptation task in a state-of-art presentation transcription RNN-T system is applied and various data filtering techniques and significantly bridge the gap between sup/unsup adaptation.
- 22
- Part-aligned pose-guided recurrent network for action recognition22
A Part-aligned Pose-guided Recurrent Network (P2RN) for action recognition is proposed, which is a recurrent network which can exploit the body representation and simultaneously model spatio-temporal evolutions of human body parts.
- Frame-GAN: Increasing the frame rate of gait videos with generative adversarial networks22
In order to increase the frame rate of gait videos, a new kind of generative adversarial networks (GAN) named Frame-GAN is proposed to reduce the gap between adjacent frames and a new effective loss function named Margin Ratio Loss (MRL) to boost the recognition model.
- Hierarchical Memory Modelling for Video Captioning22
A Hierarchical Memory Model (HMM) is proposed - a novel deep video captioning architecture which unifies a textual memory, a visual memory and an attribute memory in a hierarchical way and can largely reduce the semantic discrepancy between video and sentence.
- 20
- Generalized Inter-class Loss for Gait Recognition19
The proposed generalized inter-class loss can significantly improve the performance and achieves the state-of-the-art performances and automatically adjusts the margin between classes which enables the inter- class feature distribution to be more flexible.
- Deep Learning Model for Predicting the Pathological Complete Response to Neoadjuvant Chemoradiotherapy of Locally Advanced Rectal Cancer19
The DeepPCR model showed high accuracy in predicting pCR and served as an independent predictive factor for pCR in this study and can be used to assist in clinical treatment decision making before surgery.
- vEpiNet: A multimodal interictal epileptiform discharge detection method based on video and electroencephalogram data18
It is suggested that video data can significantly improve the performance and precision of IED detection, especially in prospective real clinic testing, and suggests that vEpiNet is a clinically viable and effective tool for IED analysis in real-world applications.
- 18
- Actor and Action Modular Network for Text-Based Video Segmentation18
A novel actor and action modular network that individually localizes the actor and its action in two separate modules is proposed that achieves the state-of-the-art performance for both single-frame segmentation and full video segmentation on A2D Sentences and J-HMDB Sentences datasets.
- 18
- 17
- Re-ranking image-text matching by adaptive metric fusion17
This paper proposes a heuristic re-ranking method called Adaptive Metric Fusion (AMF) for image-text matching that can obtain a better metric by adaptively fusing metrics based on two modules: 1) Cross-modal Reciprocal Encoding and 2) Query Replacement Gap.
- 17
- 16
- 16
- Changes in targeted metabolomics in lung tissue of chronic obstructive pulmonary disease16
The study suggested the dysregulation in amino acid metabolism, energy production pathway and perhaps lipid metabolism may be significantly related to the pathogenesis of COPD.
- 15
- 14
- EC-Flow: Enabling Versatile Robotic Manipulation from Action-Unlabeled Videos via Embodiment-Centric Flow13
Embodiment-Centric Flow (EC-Flow) is presented, a framework that directly learns manipulation from action-unlabeled videos by predicting embodimentcentric flow, and a goal-alignment module is introduced to connect the EC-Flow with language instructions and object interactions.
- 13
- 12
- 11
- 11
- 11
- Efficient Human Pose Estimation by Learning Deeply Aggregated Representations11
This paper proposes an efficient human pose estimation network (DANet) by learning deeply aggregated representations and proposes an orthogonal attention block (OAB) and a second-order fusion unit (SFU) that could achieve comparable or even better accuracy with much smaller model complexity.
- Learning a Robust Part-Aware Monocular 3D Human Pose Estimator via Neural Architecture Search9
The proposed method consists of two learning stages: searching suitable decoder architectures for specific parts and training the part-aware 3D human pose estimator built with these optimized neural architectures, which can advance state-of-the-art accuracy on both the single-person and multi-person 3Dhuman pose estimation benchmarks with affordable computational cost.
- Hierarchical Graph Convolutional Network for Skeleton-Based Action Recognition9
A novel Hierarchical Graph Convolutional Network (HiGCN) is proposed to deal with hierarchical information of human actions, which is differentiable and can be plugged into GCN in an end-to-end manner.
- 9
- 8
- Analysis of gait characteristics and related factors in patients with Parkinson's disease based on wearable devices8
It is essential to identify PIGD and develop targeted interventions to reduce the risk of PIGD in PD patients, which seriously affects patients’ quality of life.
- 8
- 8
- 8
- Object Detection and Information Perception by Fusing YOLO-SCG and Point Cloud Clustering7
This work proposes an information perception algorithm with vision as the core and the fusion of LiDAR, which is able to detect objects faster and more accurately and integrates the detection results of vision for local clustering when processing point clouds.
- 7
- 7
- 7
- Recurrent Deconvolutional Generative Adversarial Networks with Application to Video Generation7
A novel model for video generation from text descriptions is proposed, which includes a recurrent deconvolutional generative adversarial network as the generator and a 3D convolutional neural network (3D-CNN) as the discriminator.
- Person Re-identification via Learning Visual Similarity on Corresponding Patch Pairs7
A complementary similarity measure and a mutually-exclusive regulation are presented to enhance the performance of Re-ID and the best rank-1 matching rate on the VIPeR dataset is improved by 4.14%.
- Enhancing Marker Scoring Accuracy through Ordinal Confidence Modelling in Educational Assessments6
This study frames confidence estimation as a classification task: predicting whether an AES-generated score correctly places a candidate in the appropriate CEFR level using score binning, and introduces a set of novel Kernel Weighted Ordinal Categorical Cross Entropy loss functions that incorporate the ordinal structure of CEFR labels.
- Multimodal Pretrained Knowledge for Real-world Object Navigation6
This work proposes a novel vision-and-language object navigation strategy that uses multimodal pretrained knowledge as a cross-modal bridge to link semantic concepts in both images and text to improve navigation supervision at key-points and enhances robustness.
- 6
- Target-Grounded Graph-Aware Transformer for Aerial Vision-and-Dialog Navigation6
The methods of the winning entry of the AVDN Challenge in ICCV CLVL 2023 are detailed, which addresses the Aerial Navigation from Dialog History task, which requires a drone agent to associate dialog history with aerial observations to reach the destination.
- MAFormer: A New Method for Radar Reflectivity Reconstructing Using Satellite Data6
The proposed MAFormer model, a novel model for reconstructing radar reflectivity using satellite data within the Transformer framework, exhibits notable advancements when compared to state-of-the-art deep learning techniques.
- 6
- Person Re-identification by Unsupervised Color Spatial Pyramid Matching6
A novel unsupervised color spatial pyramid matching UCSPM approach for person re-identification that can well cope with the variations of viewpoint, illumination and pose through the combination of illumination invariance color feature.
- 6
- SPENet: Self-guided Prototype Enhancement Network for Few-Shot Medical Image Segmentation5
A Self-guided Prototype Enhancement Network (SPENet) is proposed, which enables multi-granularity measurement between the support and query images by simultaneously generating a global prototype and an adaptive number of local prototypes.
- The first successful expanded compassionate use of Iptacopan in a patient with paroxysmal nocturnal hemoglobinuria5
A Chinese patient with efficient treatment by Iptacopan who has been switched from previous eculizumab treatment showed normal hemolysis parameters and stable thrombophilia status and was the first case of compassionate use (CU) and expanded access (EA) to unapproved drug in China and also the first attempt of IptACopan in the real world.
- 5
- UMIGen: A Unified Framework for Egocentric Point Cloud Generation and Cross-Embodiment Robotic Imitation Learning4
This work proposes UMIGen, a unified framework that consists of a handheld data collection device that requires no visual SLAM and simultaneously records point cloud observation-action pairs, and a visibility-aware optimization mechanism that extends the DemoGen pipeline to egocentric 3D observations by generating only points within the camera's field of view.
- 3
- Color image-guided very low-resolution depth image reconstruction3
This paper proposes a network for reconstructing very low-resolution depth images with high-resolution 2D color images, mining the in-scale non-local features, cross- scale non- local features and its own priori information features of 2Dcolor and depth images in the network.
- 3
- 2
- 2
- Treatment response prediction in rectal cancer patients: A radiomics study of multimodality imaging methods2
The findings indicate that radiomics features obtained from pretreatment CT, MRI, and EUS images have the potential to accurately predict treatment response in patients with LARC.
- A Hierarchical Clustering Obstacle Detection Method Applied to RGB-D Cameras2
A fast obstacle-detection process applied to RGB-D cameras that omits the iterative computation process required by traditional methods and greatly reduces the memory and time overhead is proposed.
- 2
- 2
- 2
- 2
- PokerNet: Expanding Features Cheaply via Depthwise Convolutions2
A novel Poker module to expand features by taking advantage of cheap depthwise convolution is proposed, which can greatly reduce the computational cost, and meanwhile generate a large number of effective features to guarantee the performance.
- 2
- VicKAM: Visual Conceptual Knowledge Guided Action Map for Weakly Supervised Group Activity Recognition1
A Visual Conceptual Knowledge Guided Action Map framework that uses prototypes to produce individual action maps that indicate the likelihood of actions occurring at different locations, integrated with action semantic representations for group activity recognition.
- 1
- HCL-MTSAD: Hierarchical Contrastive Consistency Learning for Accurate Detection of Industrial Multivariate Time Series Anomalies1
A novel self-supervised hierarchical contrastive consistency learning method for detecting anomalies in MTS, named HCL-MTSAD, which innovatively leverages data consistency at multiple levels inherent in industrial MTS, systematically capturing consistent associations across four latent levels-measurement, sample, channel, and process.
- Learning to estimate optical flow using dual-frequency paradigm1
A dual-frequency encoder is proposed for optical flow estimation that captures discriminative features with both high-frequency and low-frequency biases and improves the prediction of optical flow in occluded regions by enhancing the perception of high- frequency features that further improve the robustness of the network.
- 1
- 1
- Cross-scale content-based full Transformer network with Bayesian inference for object tracking1
A cross-scale full Transformer tracker with content-based prior bias, inspired by Bayesian probabilistic formulation, to introduce long-range dependency, which enriches the likelihood representation of model, and alleviates the destruction of spatial consistency.
- 1
- Exploration on the Mixed Teaching Reform of Modern Enterprise Management Course1
This paper makes full use of the existing teacher resources and explores the ideas and methods of the mixed teaching reform by using the information teaching platform, so as to further realize the student-centered education and teaching concept and improve the teaching level and quality.
- 1
- 1
- 1
- Interactive Animation Editing Based on Sketch Interaction1
The sketch-based interface explores a point in the tradeoff between expressiveness and naturalness to provide users a natural interactive environment and validate existing efforts and provide impetus for future work in the area of natural interaction research.
- –
- –
- Does Machine Learning Improve Wind Power Forecasting? An Experimental Investigation–
A systematic experimental investigation that covers two complementary stages, i.e., wind speed correction and wind power forecasting, provides actionable task-specific guidance for model selection in operational wind power forecasting systems.
- –
- –
- –
- GVSE++: improved gated visual-semantic embedding for few-shot image and sentence matching–
A Few-Shot Visual-Semantic Embedding (FVSE) model to detect and represent few-shot instances and words, which exploits pretrained models on multiple auxiliary resources to extract their generic representations, and then associates them by performing multidirectional alignment and relation modeling to obtain their local and global similarity metrics.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Exploring the Benefits of Text and Sketch in Video Retrieval of Complex Queries–
This paper presents the first systematic comparison of these two input styles and analyzes the responses and feedbacks of users to test two different styles of inputting the semantics.
- Efficient multimedia distribution framework for Internet video–
A novel framework for Internet video streaming with 3D-SLCCA codec, which provides high coding efficiency, multiresolution video representation, transmission error resilience, and low computational complexity is proposed.
- –
- –
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 2026-10-15. 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.