Multi-label Classification on Charades
44.9mAPQ-scoring
Evaluation results
| Method | Links | |
|---|---|---|
| Q-scoringBackbone=CLIP (ViT-B/16), Aggregation=Q-scoring, Frame=32, Finetune=true2022.05 | 44.9 | |
| ActionCLIPBackbone=CLIP (ViT-B/16), Aggregation=ActionCLIP, Frame=32x10x3, Finetune=true2022.05 | 44.6 | |
| Joint. S-ABackbone=CLIP (ViT-B/16), Aggregation=Joint. S-A, Frame=32, Finetune=true2022.05 | 42.2 | |
| Temp. S-ABackbone=CLIP (ViT-B/16), Aggregation=Temp. S-A, Frame=32, Finetune=true2022.05 | 36.3 | |
| MeanBackbone=CLIP (ViT-B/16), Aggregation=Mean, Frame=32, Finetune=true2022.05 | 33 | |
| Clip4clipseqTrBackbone=CLIP (ViT-B/16), Aggregation=Clip4clipseqTr, Frame=32, Finetune=true2022.05 | 32 | |
| Q-scoringBackbone=CLIP (ViT-B/16), Aggregation=Q-scoring, Frame=32, Finetune=false2022.05 | 21.1 | |
| MeanBackbone=CLIP (ViT-B/16), Aggregation=Mean, Frame=32, Finetune=false2022.05 | 17.5 |