BEiT: BERT Pre-Training of Image Transformers
This addresses the problem of reducing reliance on labeled data for computer vision models, offering a competitive self-supervised approach for researchers and practitioners, though it builds incrementally on BERT's ideas.
The paper tackles self-supervised pre-training for vision Transformers by proposing BEiT, which uses masked image modeling to recover visual tokens from corrupted patches, achieving 83.2% top-1 accuracy on ImageNet-1K with a base model and 86.3% with a large model, outperforming previous methods like DeiT and ViT-L.
We introduce a self-supervised vision representation model BEiT, which stands for Bidirectional Encoder representation from Image Transformers. Following BERT developed in the natural language processing area, we propose a masked image modeling task to pretrain vision Transformers. Specifically, each image has two views in our pre-training, i.e, image patches (such as 16x16 pixels), and visual tokens (i.e., discrete tokens). We first "tokenize" the original image into visual tokens. Then we randomly mask some image patches and fed them into the backbone Transformer. The pre-training objective is to recover the original visual tokens based on the corrupted image patches. After pre-training BEiT, we directly fine-tune the model parameters on downstream tasks by appending task layers upon the pretrained encoder. Experimental results on image classification and semantic segmentation show that our model achieves competitive results with previous pre-training methods. For example, base-size BEiT achieves 83.2% top-1 accuracy on ImageNet-1K, significantly outperforming from-scratch DeiT training (81.8%) with the same setup. Moreover, large-size BEiT obtains 86.3% only using ImageNet-1K, even outperforming ViT-L with supervised pre-training on ImageNet-22K (85.2%). The code and pretrained models are available at https://aka.ms/beit.