Improving Vision Transformers by Revisiting High-frequency Components

Benchmark Model Rank Results
domain-generalization-on-imagenet-cVOLO-D5+HAT#10mean Corruption Error (mCE): 38.4Number of params: 296M
domain-generalization-on-imagenet-rVOLO-D5+HAT#15Top-1 Error Rate: 40.3
domain-generalization-on-stylized-imagenetVOLO-D5+HAT#2Top 1 Accuracy: 25.9
image-classification-on-imagenetVOLO-D5+HAT#97Top 1 Accuracy: 87.3%Number of params: 295.5MGFLOPs: 412