Spatio-temporal Action Localization on J-HMDB-21
93.1Video mAP (IoU=0.2)STAR/L
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| STAR/LPretraining=CLIP→K700, Backbone=ViViT/L2023.04 | 93.1 | 92.1 | 92.6 | — | |
| STAR/BPretraining=IN21K→K400, Backbone=ViViT/B2023.04 | 89.1 | 86.6 | 88.5 | — | |
| YOWO+LFBcausality=non-causal2019.11 | 88.3 | 75.7 | 85.9 | 58.6 | |
| YOWOinput_length=16-frame2019.11 | 87.8 | 74.4 | 85.7 | 58.1 | |
| TubeRPretraining=IG65M→K400, Backbone=CSN-1522023.04 | 87.4 | — | 82.3 | — | |
| P3D-CTN2019.11 | 84 | 71.1 | 80.5 | — | |
| TubeR with flowPretraining=K400, Backbone=I3D2023.04 | 81.8 | — | 80.7 | — | |
| T-CNN2019.11 | 78.4 | 61.3 | 76.9 | — | |
| MOCPretraining=IN1K → COCO, Backbone=DLA342023.04 | 77.3 | 70.8 | 77.2 | — | |
| TPnet2019.11 | 74.8 | — | 74.1 | 61.3 | |
| PengMotion Region (MR)=with2019.11 | 74.3 | 58.5 | 73.1 | — | |
| ACT2019.11 | 74.2 | 65.7 | 73.7 | 52.1 | |
| ACTPretraining=IN1K, Backbone=VGG2023.04 | 74.2 | 65.7 | 73.7 | — | |
| ROAD2019.11 | 73.8 | — | 72 | 44.5 | |
| PengMotion Region (MR)=without2019.11 | 71.1 | 56.9 | 70.6 | 48.2 | |
| WOOPretraining=K600, Backbone=SlowFast2023.04 | — | 80.5 | — | — |