Hi-EADN: Hierarchical Excitation Aggregation and Disentanglement Frameworks for Action Recognition Based on Videos.
Zeyuan HuEung-Joo LeePublished in: Symmetry (2021)
Keyphrases
- action recognition
- human actions
- video dataset
- recognition of human actions
- action classification
- spatio temporal interest points
- spatio temporal
- recognizing human actions
- ucf sports
- view invariant
- action detection
- recognizing actions
- static images
- human activities
- bag of words
- action recognition in videos
- motion features
- space time interest points
- activity recognition
- spatial temporal
- human detection
- computer vision
- mid level
- human object interactions
- depth sensors
- human pose
- bag of features
- body parts
- video sequences
- motion history images
- low level descriptors
- mid level features
- video frames
- atomic actions
- motion recognition
- human motion
- action primitives
- video content
- motion capture data
- space time
- video data