Video Classification on Kinetics
83.3Top-1 AccuracyG-Blend
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| G-BlendPre-train=IG-65M, GFLOPs=110.1x302019.05 | 83.3 | 96 | |
| ip-CSN-152Pre-train=IG-65M, GFLOPs=108.8x302019.05 | 82.5 | 95.3 | |
| G-BlendPre-train=Sports1M, GFLOPs=110.1x302019.05 | 80.4 | 94.8 | |
| SlowFast+NLPre-train=None, GFLOPs=234x302019.05 | 79.8 | 93.9 | |
| ip-CSN-152Pre-train=Sports1M, GFLOPs=108.8x302019.05 | 79.2 | 93.8 | |
| G-BlendPre-train=None, GFLOPs=110.1x302019.05 | 79.1 | 93.9 | |
| SlowFastPre-train=None, GFLOPs=213x302019.05 | 78.9 | 93.5 | |
| X-Ray VisualEncoder=EViT-2b, Protocol=Linear probing2026.02 | 78.1 | 98.37 | |
| ip-CSN-152Pre-train=None, GFLOPs=108.8x302019.05 | 77.8 | 92.8 | |
| Shift-Attn NetPre-train=ImageNet, GFLOPs=NA2019.05 | 77.7 | 93.2 | |
| PE-GEncoder=PE-G, Protocol=Linear probing2026.02 | 76.68 | 92.79 | |
| DiNO-v3Encoder=ViT-H, Protocol=Linear probing2026.02 | 70.19 | 88.36 | |
| SiGLIP 2Encoder=ViT-g/16, Protocol=Linear probing2026.02 | 51.15 | 69.96 |