HRFormer: High-ResolutionTransformerforDense Prediction

Vision Transformer (ViT) [13] shows promising performance on ImageNet classification tasks.