Improving Vision Transformers by Revisiting High-frequency Components

3 Apr 2022  ·  Jiawang Bai, Li Yuan, Shu-Tao Xia, Shuicheng Yan, Zhifeng Li, Wei Liu ·

The transformer models have shown promising effectiveness in dealing with various vision tasks. However, compared with training Convolutional Neural Network (CNN) models, training Vision Transformer (ViT) models is more difficult and relies on the large-scale training set. To explain this observation we make a hypothesis that \textit{ViT models are less effective in capturing the high-frequency components of images than CNN models}, and verify it by a frequency analysis. Inspired by this finding, we first investigate the effects of existing techniques for improving ViT models from a new frequency perspective, and find that the success of some techniques (e.g., RandAugment) can be attributed to the better usage of the high-frequency components. Then, to compensate for this insufficient ability of ViT models, we propose HAT, which directly augments high-frequency components of images via adversarial training. We show that HAT can consistently boost the performance of various ViT models (e.g., +1.2% for ViT-B, +0.5% for Swin-B), and especially enhance the advanced model VOLO-D5 to 87.3% that only uses ImageNet-1K data, and the superiority can also be maintained on out-of-distribution data and transferred to downstream tasks. The code is available at:

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Image Classification ImageNet VOLO-D5+HAT Top 1 Accuracy 87.3% # 53
Number of params 295.5M # 664
GFLOPs 412 # 382
Domain Generalization ImageNet-C VOLO-D5+HAT mean Corruption Error (mCE) 38.4 # 5
Number of params 296M # 30
Domain Generalization ImageNet-R VOLO-D5+HAT Top-1 Error Rate 40.3 # 8
Domain Generalization Stylized-ImageNet VOLO-D5+HAT Top 1 Accuracy 25.9 # 2