Better plain ViT baselines for ImageNet-1k
This provides a simpler baseline for vision tasks, though it is incremental as it modifies existing methods.
The paper tackles the problem of training Vision Transformers on ImageNet-1k without complex regularization, showing that standard data augmentation suffices and achieves 80% top-1 accuracy in under a day.
It is commonly accepted that the Vision Transformer model requires sophisticated regularization techniques to excel at ImageNet-1k scale data. Surprisingly, we find this is not the case and standard data augmentation is sufficient. This note presents a few minor modifications to the original Vision Transformer (ViT) vanilla training setting that dramatically improve the performance of plain ViT models. Notably, 90 epochs of training surpass 76% top-1 accuracy in under seven hours on a TPUv3-8, similar to the classic ResNet50 baseline, and 300 epochs of training reach 80% in less than one day.