Action Recognition on HMDB51 (Top-1 and Top-5 Accuracy)
44.75Top-1 AccVLM-RLAIF
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| VLM-RLAIFLLM Size=7B, Zero-shot=true2024.02 | 44.75 | 68.37 | |
| STSepPretrained=ImgNet, Params=11.5M, FLOPs×views=9.60 × 3 G, Frames=16×1282025.12 | 41.4 | 74.7 | |
| STSep(S1-2)Pretrained=ImgNet, Params=11.3M, FLOPs×views=9.58 × 3 G, Frames=16×1282025.12 | 40.5 | 74.3 | |
| TDBNPretrained=ImgNet, Params=11.3M, FLOPs×views=9.48 × 3 G, Frames=16×1282025.12 | 40.1 | 73.9 | |
| TKSPretrained=ImgNet, Params=11.3M, FLOPs×views=9.48 × 3 G, Frames=16×1282025.12 | 40 | 74 | |
| RSNNPretrained=ImgNet, Params=11.3M, FLOPs×views=9.48 × 3 G, Frames=16×1282025.12 | 39.9 | 73.8 | |
| TcjaPretrained=ImgNet, Params=11.3M, FLOPs×views=9.49 × 3 G, Frames=16×1282025.12 | 39.9 | 73.5 | |
| STSepPretrained=ImgNet, Params=11.5M, FLOPs×views=9.60 × 1 G, Frames=16×1282025.12 | 39.8 | 72.4 | |
| TETPretrained=ImgNet, Params=11.3M, FLOPs×views=9.48 × 3 G, Frames=16×1282025.12 | 39.6 | 74 | |
| PLIFPretrained=ImgNet, Params=11.3M, FLOPs×views=9.48 × 3 G, Frames=16×1282025.12 | 39.1 | 73.2 | |
| LLaMA-VIDLLM Size=7B, Zero-shot=true, Reproduced=true2024.02 | 38.85 | 65.27 | |
| VLM-SFTLLM Size=7B, Zero-shot=true2024.02 | 38.58 | 62.37 | |
| Video-ChatGPTLLM Size=7B, Zero-shot=true2024.02 | 37.1 | 63.97 | |
| Video-LLaVALLM Size=7B, Zero-shot=true2024.02 | 36.64 | 64.03 | |
| ER-enhanced ZSAR modelVideo=(S+Obj)+, Class=ED2021.08 | 35.3 | — | |
| E2EVideo=r(2+1)d*, Class=WN2021.08 | 32.7 | — | |
| URVideo=FV, Class=WN2021.08 | 24.4 | — | |
| TS-GCNVideo=Obj, Class=WN2021.08 | 23.2 | — | |
| ZSECOCVideo=FV, Class=WN2021.08 | 22.6 | — | |
| ASRVideo=C3D*, Class=WT2021.08 | 21.8 | — | |
| STSepPretrained=false, Params=11.5M, FLOPs×views=9.60 × 3 G, Frames=16×1282025.12 | 21.2 | 50.9 | |
| STSepPretrained=false, Params=11.5M, FLOPs×views=9.60 × 3 G, Frames=8×1282025.12 | 20.8 | 50.9 | |
| MTEVideo=FV, Class=WN2021.08 | 19.7 | — | |
| TDBNPretrained=false, Params=11.3M, FLOPs×views=9.48 × 3 G, Frames=16×1282025.12 | 19.1 | 49.5 | |
| TETPretrained=false, Params=11.3M, FLOPs×views=9.48 × 3 G, Frames=16×1282025.12 | 18.7 | 47.2 | |
| ESZSLVideo=FV, Class=WN2021.08 | 18.5 | — | |
| TSNPretrained=false, Params=11.3M, FLOPs×views=9.48 × 3 G, Frames=16×1282025.12 | 17.2 | 48.2 | |
| O2AVideo=Obj, Class=WN2021.08 | 15.6 | — | |
| SJEVideo=FV, Class=WN2021.08 | 13.3 | — | |
| SVEVideo=BoW, Class=WN2021.08 | 13 | — |