DirecFormer: A Directed Attention in Transformer Approach to Robust Action Recognition.
Thanh-Dat TruongQuoc-Huy BuiChi Nhan DuongHan-Seok SeoSon Lam PhungXin LiKhoa LuuPublished in: CoRR (2022)
Keyphrases
- action recognition
- view invariant
- human actions
- bag of words
- human detection
- activity recognition
- spatial temporal
- action classification
- body parts
- computer vision
- recognizing human actions
- recognizing actions
- video dataset
- recognition of human actions
- mid level
- bag of features
- robust tracking
- independent subspace analysis
- depth sensors
- action primitives
- view invariant action recognition
- static images
- partial occlusion
- multi class
- keywords