Procedural Activities Classification on COIN
90AccuracyDist.Sup.
Evaluation results
| Method | Links | |
|---|---|---|
| Dist.Sup.#Data=136M2022.12 | 90 | |
| TranS4mer#Data=495K2022.12 | 89.23 | |
| TimeSformerPre-training Dataset=HowTo100M [36], wikiHow [29], #Training Samples=100M, Domain=Instructional2022.10 | 88.9 | |
| ViS4mer#Data=495K2022.12 | 88.41 | |
| LF-VILAPre-training Dataset=LF-VILA-8M, #Training Samples=8M, Domain=Open2022.10 | 85.7 | |
| TimeSformerPre-training Dataset=HowTo 100M [36], #Training Samples=100M, Domain=Instructional2022.10 | 85.3 | |
| TimeSformerPre-training Dataset=Kinetics [7], #Training Samples=370K, Domain=Action2022.10 | 83.5 | |
| TSN#Data=306K2022.12 | 73.4 | |
| VideoCLIPPre-training Dataset=HowTo 100M [36], #Training Samples=100M, Domain=Instructional2022.10 | 72.5 | |
| SlowFastPre-training Dataset=Kinetics [7], #Training Samples=370K, Domain=Action2022.10 | 71.6 | |
| MIL-NCEPre-training Dataset=HowTo100M [36], #Training Samples=100M, Domain=Instructional2022.10 | 70.2 | |
| ClipBERTPre-training Dataset=COCO [9], Visual Genome [24], #Training Samples=5.6M, Domain=Open2022.10 | 65.4 |