AttT2M: Text-Driven Human Motion Generation with Multi-Perspective Attention Mechanism.
Chongyang ZhongLei HuZihao ZhangShihong XiaPublished in: ICCV (2023)
Keyphrases
- human motion
- multi perspective
- attention mechanism
- motion capture
- human body
- image sequences
- optical motion capture
- d scene
- video sequences
- human actions
- motion recognition
- motion capture data
- spatio temporal
- visual attention
- multi modal
- visual attention model
- human movements
- object tracking
- saliency map
- feature space
- keywords