Action Recognition on Kinetics 200 (180/20)
84.6Top-1 AccuracyTResNet-L
Evaluation Results
| Method | Links | |
|---|---|---|
| TResNet-LPre-training Dataset=ImageNet-21K-P2021.04 | 84.6 | |
| TResNet-MPre-training Dataset=ImageNet-21K-P2021.04 | 84.3 | |
| TResNet-LPre-training Dataset=ImageNet-1K2021.04 | 82.1 | |
| ResNet50Pre-training Dataset=ImageNet-21K-P2021.04 | 81.3 | |
| TResNet-MPre-training Dataset=ImageNet-1K2021.04 | 80.5 | |
| ResNet50Pre-training Dataset=ImageNet-1K2021.04 | 78.2 | |
| OFA595Pre-training Dataset=ImageNet-21K-P2021.04 | 78.1 | |
| MobileNetV3Pre-training Dataset=ImageNet-21K-P2021.04 | 74.3 | |
| OFA595Pre-training Dataset=ImageNet-1K2021.04 | 73.2 | |
| MobileNetV3Pre-training Dataset=ImageNet-1K2021.04 | 72.2 | |
| TDSMProtocol=Zero-shot, Text descriptions=12024.11 | 38.18 | |
| PURLSProtocol=Zero-shot, Text descriptions=72024.11 | 32.22 | |
| PURLSProtocol=Zero-shot, Text descriptions=12024.11 | 25.96 | |
| ReViSEProtocol=Zero-shot2024.11 | 24.95 | |
| DeViSEProtocol=Zero-shot2024.11 | 22.22 |