Action Recognition on Charades-Ego first-person (test)
0.338mAPHierVL-SA
Evaluation Results
| Method | Links | |
|---|---|---|
| HierVL-SAEvaluation Protocol=Fine-tuned2023.01 | 0.338 | |
| HierVL-w/o HierEvaluation Protocol=Fine-tuned2023.01 | 0.326 | |
| HierVL-AvgEvaluation Protocol=Fine-tuned2023.01 | 0.326 | |
| Frozen+EgoNCEVis Enc=TimeSformer, # Frames=16, Vis-Text PT=EgoClip, Train / FT Data=Charades-Ego (1st), Zero-shot evaluation=false2022.06 | 0.321 | |
| EgoVLPEvaluation Protocol=Fine-tuned2023.01 | 0.321 | |
| FrozenVis Enc=TimeSformer, # Frames=16, Vis-Text PT=EgoClip, Train / FT Data=Charades-Ego (1st), Zero-shot evaluation=false2022.06 | 0.312 | |
| FrozenVis Enc=TimeSformer, # Frames=16, Vis-Text PT=CC3M+WebVid-2M, Train / FT Data=Charades-Ego (1st), Zero-shot evaluation=false2022.06 | 0.309 | |
| Ego-ExoVis Enc=SlowFast (ResNet-101), # Frames=32, Vis-Text PT=-, Train / FT Data=Charades-Ego (1st), Zero-shot evaluation=false2022.06 | 0.301 | |
| Ego-ExoEvaluation Protocol=Fine-tuned2023.01 | 0.301 | |
| FrozenVis Enc=TimeSformer, # Frames=16, Vis-Text PT=-, Train / FT Data=Charades-Ego (1st), Zero-shot evaluation=false2022.06 | 0.288 | |
| FrozenVis Enc=TimeSformer, # Frames=16, Vis-Text PT=HowTo100M, Train / FT Data=Charades-Ego (1st), Zero-shot evaluation=false2022.06 | 0.283 | |
| I3DVis Enc=13D, # Frames=32, Vis-Text PT=-, Train / FT Data=Charades-Ego (1st), Zero-shot evaluation=false2022.06 | 0.258 | |
| I3DEvaluation Protocol=Fine-tuned2023.01 | 0.258 | |
| Frozen+EgoNCEVis Enc=TimeSformer, # Frames=16, Vis-Text PT=EgoClip, Train / FT Data=-, Zero-shot evaluation=true2022.06 | 0.25 | |
| FrozenVis Enc=TimeSformer, # Frames=16, Vis-Text PT=EgoClip, Train / FT Data=-, Zero-shot evaluation=true2022.06 | 0.236 | |
| SSDAVis Enc=13D, # Frames=32, Vis-Text PT=-, Train / FT Data=Charades-Ego (1st + 3rd), Zero-shot evaluation=false2022.06 | 0.231 | |
| SSDAEvaluation Protocol=Fine-tuned2023.01 | 0.231 | |
| FrozenVis Enc=TimeSformer, # Frames=16, Vis-Text PT=CC3M+WebVid-2M, Train / FT Data=-, Zero-shot evaluation=true2022.06 | 0.209 | |
| ActorVis Enc=ResNet-152, # Frames=25, Vis-Text PT=-, Train / FT Data=Charades-Ego (1st + 3rd), Zero-shot evaluation=false2022.06 | 0.2 | |
| ActorEvaluation Protocol=Fine-tuned2023.01 | 0.2 | |
| FrozenVis Enc=TimeSformer, # Frames=16, Vis-Text PT=HowTo100M, Train / FT Data=-, Zero-shot evaluation=true2022.06 | 0.092 |