Action Recognition on Assembly101 (test)
62.7Top-1 AccuracyTimeSformer
Evaluation Results
| Method | Links | |
|---|---|---|
| TimeSformerInference Mode=Exocentric, Distillation=False2026.06 | 62.7 | |
| UNIEGOInference Mode=Egocentric, Distillation=True2026.06 | 50.7 | |
| Multiteacher Dist.Inference Mode=Egocentric, Distillation=True2026.06 | 48.2 | |
| π-ViTInference Mode=Egocentric, Distillation=True2026.06 | 47.8 | |
| TimeSformerInference Mode=Egocentric, Distillation=False2026.06 | 47.6 | |
| ViFi-CLIPInference Mode=Egocentric, Distillation=False2026.06 | 46.6 | |
| ST-GCNInference Mode=Exocentric, Distillation=False2026.06 | 46.2 | |
| eSEC–LAMcategory=Proposed Method2026.04 | 43 | |
| HandFormercategory=End-to-End Video Model2026.04 | 41.1 | |
| TSM+TAInference Mode=Egocentric, Distillation=False2026.06 | 40.5 | |
| TSMcategory=End-to-End Video Model2026.04 | 33.8 | |
| Classical eSECstatus=re-implemented2026.04 | 31.5 | |
| MS-G3Dcategory=End-to-End Video Model2026.04 | 28.7 |