Objectron: A Large Scale Dataset of Object-Centric Videos in the Wild With Pose Annotations.
Adel AhmadyanLiangkai ZhangArtsiom AblavatskiJianing WeiMatthias GrundmannPublished in: CVPR (2021)
Keyphrases
- d objects
- trecvid multimedia event detection
- annotation tool
- human actions
- object representation
- million images
- event recognition
- event detection
- pose estimation
- position and orientation
- video sequences
- object geometry
- partial occlusion
- human pose
- user generated
- pose recovery
- natural language descriptions
- object pose
- multimedia event detection
- weakly labeled
- video collections
- video dataset
- moving objects
- articulated objects
- object tracking
- multiple objects
- object model
- video frames
- semantic annotation
- video clips
- action recognition
- crowd sourced
- video content
- object categories
- object shape
- web videos
- object appearance
- object models
- rigid objects