Video Recognition on Kinetics-400 fully-supervised
82.3Top-1 AccuracyXCLIP
Evaluation results
| Method | Links | ||
|---|---|---|---|
| XCLIPFrames=8, Tunable Param=132, single-view inference=true2024.11 | 82.3 | 95.8 | |
| ViCLIP + MSTA+LCCFrames=8, Tunable Param=8.7, single-view inference=true2024.11 | 82.2 | 96.2 | |
| OSTFrames=8, Tunable Param=149.6, single-view inference=true2024.11 | 82 | 95.8 | |
| ViCLIP + MSTAFrames=8, Tunable Param=8.7, single-view inference=true2024.11 | 81.6 | 95.8 | |
| Vita-CLIPFrames=8, Tunable Param=39, single-view inference=true2024.11 | 80.5 | 95.9 | |
| ViCLIP + LORAFrames=8, Tunable Param=9.4, single-view inference=true2024.11 | 80.3 | 95.4 | |
| ViCLIP + AdaptFormerFrames=8, Tunable Param=7.9, single-view inference=true2024.11 | 80.2 | 95.5 | |
| ViCLIPFrames=8, Tunable Param=124.3, single-view inference=true2024.11 | 79.9 | 95.1 |