2021/05/29 Updated by
ROSE: Rapid-Rich Object Search Lab
[Up]
Action Recognition Datasets: "NTU RGB+D" and "NTU RGB+D 120"
- official site
- github
- NTU RGB+D: 60 動作クラス+ 56,880 動画
- NTU RGB+D 120: 別の60 動作クラス+ 57,600 動画 (合計: 120動作クラス + 114,480動画)
- RGB video: 1920x1080
- depth map and IR video: 512x424
- 3D skeleton data: 3D coordinates of 25 joints