Image Classification on Avg 5 (ImageNet, CIFAR-10, CIFAR-100, Pets, Flowers) 1 (test)
95.97Top-1 AccuracyViT-H/14
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| ViT-H/14Pre-training dataset=JFT-300M (proprietary), Resolution=448, Evaluation protocol=Fine-tuned2021.05 | 95.97 | 15 | 2,300 | |
| Mixer-H/14Pre-training dataset=JFT-300M (proprietary), Resolution=448, Evaluation protocol=Fine-tuned2021.05 | 95.71 | 40 | 1,010 | |
| BiT-R152x4Pre-training dataset=JFT-300M (proprietary), Resolution=448, Evaluation protocol=Fine-tuned2021.05 | 95.33 | 26 | 9,900 | |
| ViT-L/16Pre-training dataset=ImageNet-21k (public), Resolution=448, Evaluation protocol=Fine-tuned2021.05 | 94.39 | 32 | 180 | |
| BiT-R152x4Pre-training dataset=ImageNet-21k (public), Resolution=448, Evaluation protocol=Fine-tuned2021.05 | 94.04 | 26 | 940 | |
| Mixer-L/16Pre-training dataset=ImageNet-21k (public), Resolution=448, Evaluation protocol=Fine-tuned2021.05 | 93.91 | 105 | 410 |