Text-to-Video Retrieval on MSRVTT 1K 1.0 (test)
40.9R@1OA-Trans
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| OA-TransVis Enc. Init.=CLIP-WIT, Pretrained Data=[5.5M] CC3M, WebVid-2M, Number of Frames=12, Zero-shot=false2021.12 | 40.9 | 70.4 | 80.3 | 2 | |
| OA-TransVis Enc. Init.=CLIP-WIT, Pretrained Data=[5.5M] CC3M, WebVid-2M, Zero-shot=false2021.12 | 39.4 | 68.8 | 78.3 | 2 | |
| OA-TransVis Enc. Init.=ImageNet, Pretrained Data=[5.5M] CC3M, WebVid-2M, Zero-shot=false2021.12 | 35.8 | 63.4 | 76.5 | 3 | |
| Frozen (Our Imp.)Year=ICCV'21, Vis Enc. Init.=ImageNet, Pretrained Data=[5.5M] CC3M, WebVid-2M, Zero-shot=false2021.12 | 33.2 | 61.5 | 71.9 | 3 | |
| OA-TransVis Enc. Init.=ImageNet, Pretrained Data=[2.5M] Webvid-2M, Zero-shot=false2021.12 | 32.7 | 60.9 | 72.5 | 3 | |
| OA-TransVis Enc. Init.=CLIP-WIT, Pretrained Data=[5.5M] CC3M, WebVid-2M, Number of Frames=12, Zero-shot=true2021.12 | 31.4 | 55.3 | 64.8 | 4 | |
| FrozenYear=ICCV'21, Vis Enc. Init.=ImageNet, Pretrained Data=[5.5M] CC3M, WebVid-2M, Zero-shot=false2021.12 | 31 | 59.5 | 70.5 | 3 | |
| Support SetYear=ICLR'21, Vis Enc. Init.=IG65M, ImageNet, Pretrained Data=[136M] HowTo100M, Zero-shot=false2021.12 | 30.1 | 58.5 | 69.3 | 3 | |
| OA-TransVis Enc. Init.=CLIP-WIT, Pretrained Data=[5.5M] CC3M, WebVid-2M, Zero-shot=true2021.12 | 29.7 | 52.1 | 63.5 | 5 | |
| CLIPYear=Arxiv'21, Vis Enc. Init.=CLIP-WIT, Number of Frames=12, Zero-shot=true2021.12 | 28.5 | 49.7 | 61.2 | 5 | |
| MMTYear=ECCV'20, Vis Enc. Init.=Numerous experts, Pretrained Data=[136M] HowTo100M, Zero-shot=false2021.12 | 26.6 | 57.1 | 69.6 | 4 | |
| FrozenYear=ICCV'21, Vis Enc. Init.=ImageNet, Pretrained Data=[3M] CC3M, Zero-shot=false2021.12 | 25.5 | 54.5 | 66.1 | 4 | |
| OA-TransVis Enc. Init.=ImageNet, Pretrained Data=[5.5M] CC3M, WebVid-2M, Zero-shot=true2021.12 | 23.4 | 47.5 | 55.6 | 8 | |
| ClipBERTYear=ICCV'21, Pretrained Data=[5.6M] COCO, VisGenome, Zero-shot=false2021.12 | 22 | 46.8 | 59.9 | 6 | |
| Frozen [Our Imp.]Year=ICCV'21, Vis Enc. Init.=ImageNet, Pretrained Data=[5.5M] CC3M, WebVid-2M, Zero-shot=true2021.12 | 21.7 | 45.5 | 53.9 | 9 | |
| FrozenYear=ICCV'21, Vis Enc. Init.=ImageNet, Pretrained Data=[5.5M] CC3M, WebVid-2M, Zero-shot=true2021.12 | 18.7 | 39.5 | 51.6 | 10 | |
| OA-TransVis Enc. Init.=ImageNet, Pretrained Data=[2.5M] WebVid-2M, Zero-shot=true2021.12 | 18.4 | 36.5 | 46.8 | 10 | |
| NEYear=AAAI'21, Vis Enc. Init.=ImageNet, Kinetics, Pretrained Data=[136M] HowTo100M, Zero-shot=false2021.12 | 17.4 | 41.6 | 53.6 | 8 | |
| ActBERTYear=CVPR'20, Vis Enc. Init.=VisGenome, Pretrained Data=[136M] HowTo100M, Zero-shot=false2021.12 | 16.3 | 42.8 | 56.9 | 10 | |
| VidTranslateYear=Arxiv'20, Vis Enc. Init.=IG65M, Pretrained Data=[136M] HowTo100M, Zero-shot=false2021.12 | 14.7 | — | 52.8 | — | |
| FrozenYear=ICCV'21, Vis Enc. Init.=ImageNet, Pretrained Data=[2.5M] WebVid-2M, Zero-shot=true2021.12 | 14.5 | 29.5 | 64.5 | 21 | |
| SupportSetYear=ICLR'21, Vis Enc. Init.=IG65M, ImageNet, Pretrained Data=[136M] HowTo100M, Zero-shot=true2021.12 | 8.7 | 23 | 31.1 | 31 | |
| HT MIL-NCEYear=CVPR'20, Pretrained Data=[136M] HowTo100M, Zero-shot=true2021.12 | 7.5 | 21.2 | 29.6 | 38 |