Image Representation Learning on ImageNet 1k (train)
78.9AccuracyMAGE VIT-L
Evaluation results
| Method | Links | |
|---|---|---|
| MAGE VIT-LType=enc.-dec., number of tokens=256, number of parameters=404M2023.12 | 78.9 | |
| VIM+ViTType=dec.-only, number of tokens=1024, number of parameters=650M2023.12 | 65.1 | |
| GIVT-CausalType=dec.-only, number of tokens=256, number of parameters=304M2023.12 | 65.1 | |
| VIM+CNNType=dec.-only, number of tokens=1024, number of parameters=650M2023.12 | 61.8 | |
| BigBiGANnumber of parameters=344M2023.12 | 61.3 | |
| iGPT-LType=dec.-only, number of tokens=1024, number of parameters=1362M2023.12 | 60.3 |