A5065122626- Human Pose and Action Recognition
- Anomaly Detection Techniques and Applications
- Gait Recognition and Analysis
- Video Surveillance and Tracking Methods
Xi'an Jiaotong University
2019-2020
Skeleton-based action recognition has achieved great advances with the development of graph convolutional networks (GCNs). Many existing GCNs-based models only use fixed hand-crafted adjacency matrix to describe connections between human body joints. This omits important implicit joints, which contain discriminative information for different actions. In this paper, we propose an action-specific module, is able extract and properly balance them each action. addition, filter out useless...
Graph structure is an important part of convolutional networks (GCNs), which can reflect the connection between each nodes non-Euclidean data. A feature hidden in graph structure, provide additional spatial features that represent relationship human joints. However many GCNs-based methods ignore these features. We put forward a extraction module, obtain implicit joints, and extract from structural In order to enhance temporal representation, we propose long-range frame-difference module....