A5045183950- Human Pose and Action Recognition
- Advanced Vision and Imaging
- Video Surveillance and Tracking Methods
- Hand Gesture Recognition Systems
- Advanced Image and Video Retrieval Techniques
- ECG Monitoring and Analysis
- Image Processing Techniques and Applications
- Robot Manipulation and Learning
- EEG and Brain-Computer Interfaces
- Face recognition and analysis
- Advanced Neural Network Applications
- Optical measurement and interference techniques
- Video Coding and Compression Technologies
- Thermochemical Biomass Conversion Processes
- Advanced Image Processing Techniques
- Face and Expression Recognition
- Robotics and Sensor-Based Localization
- Anomaly Detection Techniques and Applications
- Phonocardiography and Auscultation Techniques
- Advanced Data Compression Techniques
- Lignin and Wood Chemistry
- Visual Attention and Saliency Detection
- Gait Recognition and Analysis
- Image Enhancement Techniques
- Image Retrieval and Classification Techniques
Tsinghua University
2016-2025
Beijing Academy of Artificial Intelligence
2023-2024
Shanghai Artificial Intelligence Laboratory
2022-2024
Dalian University
2020-2023
Dalian University of Technology
2020-2023
University Town of Shenzhen
2021-2023
Dalian Institute of Chemical Physics
2017-2021
National Engineering Research Center for Information Technology in Agriculture
2020-2021
Chinese Academy of Sciences
2017-2021
Shandong University of Science and Technology
2021
In this paper, we strive to answer two questions: What is the current state of 3D hand pose estimation from depth images? And, what are next challenges that need be tackled? Following successful Hands Million Challenge (HIM2017), investigate top 10 state-of-the-art methods on three tasks: single frame estimation, tracking, and during object interaction. We analyze performance different CNN structures with regard shape, joint visibility, view point articulation distributions. Our findings...
Hand pose estimation from monocular depth images is an important and challenging problem for human-computer interaction. Recently deep convolutional networks (ConvNet) with sophisticated design have been employed to address it, but the improvement over traditional methods not so apparent. To promote performance of directly 3D coordinate regression, we propose a tree-structured Region Ensemble Network (REN), which partitions convolution outputs into regions integrates results multiple...
Owing to visual ambiguities and disparities, person re-identification methods inevitably produce sub optimal rank-list, which still requires exhaustive human eyeballing identify the correct target from hundreds of different likely-candidates. Existing studies focus on improving ranking performance, but rarely look into critical problem optimising time-consuming error-prone post-rank search at user end. In this study, we present a novel one-shot Post-rank Optimization (POP) method, allows...
3-D hand pose estimation is an essential problem for human-computer interaction. Most of the existing depth-based methods consume 2-D depth map or volume via 2-D/3-D convolutional neural networks. In this paper, we propose a deep semantic regression network (SHPR-Net) from point sets, which consists two subnetworks: segmentation subnetwork and subnetwork. The assigns labels each in set. integrates priors with both input late fusion strategy regresses final pose. Two transformation matrices...
Dynamic hand gesture recognition has attracted increasing interests because of its importance for human computer interaction. In this paper, we propose a new motion feature augmented recurrent neural network skeleton-based dynamic recognition. Finger features are extracted to describe finger movements and global utilized represent the movement skeleton. These then fed into bidirectional (RNN) along with skeleton sequence, which can augment RNN improve classification performance. Experiments...
In this paper, we propose a novel patient-specific electrocardiogram (ECG) classification algorithm based on the recurrent neural networks (RNN) and density clustering technique. We use RNN to learn time correlation among ECG signal points classify beats with different heart rates. Morphology information including present beat T wave of former is fed into underlying features automatically. Clustering method employed find representative as training data. Evaluated MIT-BIH Arrhythmia Database,...
Exploiting both RGB (2D appearance) and Depth (3D geometry) information can improve the performance of semantic segmentation. However, due to inherent difference between information, it remains a challenging problem in how integrate RGB-D features effectively. In this letter, address issue, we propose Non-local Aggregation Network (NANet), with well-designed Multi-modality Module (MNAM), better exploit non-local context at multi-stage. Compared most existing segmentation schemes, which only...
Graph convolutional networks (GCN) have recently been studied to exploit the graph topology of human body for skeleton-based action recognition. However, most these methods unfortunately aggregate messages via an inflexible pattern various samples, lacking awareness intra-class variety and suitableness skeleton sequences, which often contain redundant or even detrimental connections. In this paper, we propose a novel Deformable Convolutional Network (DeGCN) adaptively capture informative...
Dynamic hand gesture recognition has attracted increasing attention because of its importance for human⁻computer interaction. In this paper, we propose a novel motion feature augmented network (MFA-Net) dynamic from skeletal data. MFA-Net exploits features finger and global movements to augment deep recognition. To describe articulated movements, are extracted the skeleton sequence via variational autoencoder. Global utilized represent skeleton. These along with then fed into three branches...
This paper proposes a novel high-accuracy stereo matching scheme based on adaptive ground control points (AdaptGCP). Different from traditional fixed GCP-based methods, we consider color dissimilarity, spatial relation, and the pixel-matching reliability to select GCP adaptively in each local support window. To minimize global energy, propose practical solution, named as alternating updating of disparity confidence map, which can effectively eliminate redundant interfering information...
3D hand pose estimation from single depth image is an important and challenging problem for human-computer interaction. Recently deep convolutional networks (ConvNet) with sophisticated design have been employed to address it, but the improvement over traditional random forest based methods not so apparent. To exploit good practice promote performance estimation, we propose a tree-structured Region Ensemble Network (REN) directly coordinate regression. It first partitions last convolution...
Transparent and reflective objects are omnipresent in our daily life, but their unique visual optical characteristics notoriously challenging even for state-of-the-art deep networks of semantic segmentation. To alleviate this challenge, we construct a new large-scale real-world RGB-D dataset called TROSD, which is more comprehensive than existing datasets transparent object Our TROSD contains 11,060 images with three classes terms objects, others, covering variety scenes. Together the...
In deep face recognition, the commonly used softmax loss and its newly proposed variations are not yet sufficiently effective to handle class imbalance saturation issues during training process while extracting discriminative features. this brief, address both issues, we propose a class-variant margin (CVM) normalized loss, by introducing true-class false-class into cosine space of angle between feature vector class-weight vector. The alleviates problem, postpones early individual softmax....
Airflow patterns are essential for heating, ventilation and air conditioning (HVAC) systems. Traditional HVAC systems predesigned operated using a fixed airflow pattern. However, the indoor occupancy heat source always vary therefore, flow pattern cannot efficiently maintain required environment conditions. In this study, novel Adjustable Fan Network (AFN) improving manoeuvrability is proposed. It integrates multiple small adjustable axial fans into an AFN, enabling it to change based on...