Text-to-Video Retrieval on LSMDC (val)
0.178R@1WebVid2M
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| WebVid2MVideo-Text Training Data=WebVid2M [4], Evaluation Protocol=Zero-shot2023.10 | 0.178 | 0.312 | 0.398 | 25 | |
| HowToCaptionVideo-Text Training Data=HowToCaption (ours), Evaluation Protocol=Zero-shot2023.10 | 0.173 | 0.317 | 0.386 | 29 | |
| VideoCC3MVideo-Text Training Data=VideoCC3M [39], Evaluation Protocol=Zero-shot2023.10 | 0.148 | 0.294 | 0.358 | 33 | |
| BLIP InitializationVideo-Text Training Data=None (BLIP initialization), Evaluation Protocol=Zero-shot2023.10 | 0.147 | 0.295 | 0.365 | 31 | |
| HowTo100M (ASRs)Video-Text Training Data=HowTo100M with ASRs [17], Evaluation Protocol=Zero-shot2023.10 | 0.129 | 0.247 | 0.315 | 56 | |
| HowTo100M (distant supervision)Video-Text Training Data=HowTo100M with distant supervision [31], Evaluation Protocol=Zero-shot2023.10 | 0.121 | 0.247 | 0.324 | 42.5 | |
| HTM-AAVideo-Text Training Data=HTM-AA (auto-aligned ASRs) [17], Evaluation Protocol=Zero-shot2023.10 | 0.119 | 0.239 | 0.305 | 46 |