Track2Act: Predicting Point Tracks from Internet Videos enables Diverse Zero-shot Robot Manipulation.
Homanga BharadhwajRoozbeh MottaghiAbhinav GuptaShubham TulsianiPublished in: CoRR (2024)
Keyphrases
- manipulation tasks
- mobile robot
- video sequences
- vision system
- internet users
- real time
- robot navigation
- autonomous robots
- humanoid robot
- video frames
- robotic systems
- obstacle avoidance
- robot arm
- video content
- wide variety
- video sharing
- real world
- robot manipulators
- event recognition
- video event detection
- position and orientation
- video clips
- optical flow
- moving objects
- image sequences
- multimedia