Hierarchical Compositional Representations for Few-shot Action Recognition.
Changzhen LiJie ZhangShuzhe WuXin JinShiguang ShanPublished in: CoRR (2022)
Keyphrases
- action recognition
- mid level
- bag of words
- human actions
- activity recognition
- action classification
- human detection
- recognition of human actions
- computer vision
- body parts
- video sequences
- depth sensors
- view invariant
- recognizing human actions
- video dataset
- static images
- spatial temporal
- video data
- visual features
- action primitives
- recognizing actions
- independent subspace analysis
- view invariant action recognition
- human pose
- video shots
- video content
- human activities
- detection algorithm
- low level
- optical flow