Video Recognition on HMDB51 (test)
61.4AccuracyMoTE
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MoTEEncoder=ViT-L/14, Frames=8, Protocol=Zero-shot2024.10 | 61.4 | — | — | |
| OTIEncoder=ViT-L/14, Frames=8, Protocol=Zero-shot2024.10 | 59.3 | — | — | |
| Open-VCLIPEncoder=ViT-L/14, Frames=8, Protocol=Zero-shot2024.10 | 59 | — | — | |
| DiSTEncoder=ViT-L/14, Frames=32, Protocol=Zero-shot2024.10 | 57.5 | — | — | |
| MoTEEncoder=ViT-B/16, Frames=8, Protocol=Zero-shot2024.10 | 55.8 | — | — | |
| OTIEncoder=ViT-B/16, Frames=8, Protocol=Zero-shot2024.10 | 54.2 | — | — | |
| Open-VCLIPEncoder=ViT-B/16, Frames=8, Protocol=Zero-shot2024.10 | 53.9 | — | — | |
| Text4VisEncoder=ViT-L/14, Frames=8, Protocol=Zero-shot2024.10 | 52.4 | — | — | |
| MAXIEncoder=ViT-B/16, Frames=16/32, Protocol=Zero-shot2024.10 | 52.3 | — | — | |
| ST-AdapterEncoder=ViT-B/16, Frames=8, Protocol=Zero-shot2024.10 | 51.5 | — | — | |
| ViFi-CLIPEncoder=ViT-B/16, Frames=32, Protocol=Zero-shot2024.10 | 51.3 | — | — | |
| Text4VisZero-shot=true2022.07 | 49.8 | — | — | |
| Vita-CLIPEncoder=ViT-B/16, Frames=8/32, Protocol=Zero-shot2024.10 | 48.6 | — | — | |
| X-FlorenceEncoder=Florence, Frames=32, Protocol=Zero-shot2024.10 | 48.4 | — | — | |
| X-CLIPEncoder=ViT-B/16, Frames=32, Protocol=Zero-shot2024.10 | 44.6 | — | — | |
| A5Encoder=ViT-B/16, Frames=32, Protocol=Zero-shot2024.10 | 44.3 | — | — | |
| ActionCLIPEncoder=ViT-B/16, Frames=32, Protocol=Zero-shot2024.10 | 40.8 | — | — | |
| ResTZero-shot=true2022.07 | 34.4 | — | — | |
| E2EZero-shot=true2022.07 | 24.8 | — | — | |
| AdaptFormer-64Avg. Params (M)=1.26, Pre-training=Supervised2022.05 | — | 73.21 | — | |
| ADMM-based TT-format LSTMComp. Method=TT, # Para.=0.20M2021.07 | — | 64.09 | 84 | |
| Full-tuningAvg. Params (M)=86.04, Pre-training=Supervised2022.05 | — | 59.38 | — | |
| LinearAvg. Params (M)=0.07, Pre-training=Supervised2022.05 | — | 70.31 | — | |
| PSTRN-MComp. Method=TR, # Para.=0.36M2021.07 | — | 59.67 | 46.7 | |
| PSTRN-SComp. Method=TR, # Para.=0.48M2021.07 | — | 60.04 | 34.7 | |
| TR-LSTMComp. Method=TR, # Para.=0.67M2021.07 | — | 63.8 | 25 | |
| TT-LSTMComp. Method=TT, # Para.=0.67M2021.07 | — | 62.24 | 25 | |
| UncompressedComp. Method=-, # Para.=16.8M2021.07 | — | 62.9 | 1 | |
| VPTAvg. Params (M)=0.08, Pre-training=Supervised2022.05 | — | 71.56 | — |