Training data-efficient image transformers & distillation through attention

Code

• github.com/facebookresearch/deit
  OfficialIn paperpytorch★ 4,327
• github.com/huggingface/transformers
  pytorch★ 158,292
• github.com/rwightman/pytorch-image-models
  In paperpytorch★ 36,538
• github.com/PaddlePaddle/PaddleClas
  paddle★ 5,788
• github.com/hustvl/vim
  pytorch★ 3,823
• github.com/alibaba/EasyCV
  pytorch★ 1,949
• github.com/open-edge-platform/training_extensions
  pytorch★ 1,220
• github.com/jacobgil/vit-explain
  pytorch★ 1,074
• github.com/open-edge-platform/geti
  pytorch★ 467
• github.com/TACJu/TransFG
  pytorch★ 421

Abstract

Recently, neural networks purely based on attention were shown to address image understanding tasks such as image classification. However, these visual transformers are pre-trained with hundreds of millions of images using an expensive infrastructure, thereby limiting their adoption. In this work, we produce a competitive convolution-free transformer by training on Imagenet only. We train them on a single computer in less than 3 days. Our reference vision transformer (86M parameters) achieves top-1 accuracy of 83.1% (single-crop evaluation) on ImageNet with no external data. More importantly, we introduce a teacher-student strategy specific to transformers. It relies on a distillation token ensuring that the student learns from the teacher through attention. We show the interest of this token-based distillation, especially when using a convnet as a teacher. This leads us to report results competitive with convnets for both Imagenet (where we obtain up to 85.2% accuracy) and when transferring to other tasks. We share our code and models.

Tasks

Benchmark Results

Reproductions