ActAR: Actor-Driven Pose Embeddings for Video Action Recognition.
Soufiane LamghariGuillaume-Alexandre BilodeauNicolas SaunierPublished in: CVPR Workshops (2022)
Keyphrases
- action recognition
- action detection
- human actions
- action classification
- human pose
- recognizing actions
- spatial temporal
- video dataset
- static images
- recognizing human actions
- recognition of human actions
- human activities
- bag of words
- motion features
- human detection
- body parts
- activity recognition
- computer vision
- space time interest points
- pose estimation
- view invariant
- mid level
- human object interactions
- video sequences
- video surveillance
- video content
- video data
- video analysis
- depth sensors
- space time
- max margin
- video images
- bag of features
- video clips
- human body
- action recognition in videos
- d objects