Dat of deep learning

This article yes 2022 year Vision-Transformer In the field of CVPR The paper . The article introduces a new Vision Transformer Model ——Deformable Attention Transformer(DAT).DAT take DCN Apply to Transformer in , Thus, the operation of attention is focused on the important area , It brings a certain degree of performance improvement to target detection .DAT The greatest value is that it enables the self attention layer to Focus on relevant areas to capture information .

Refer to the directory ：
① Source code
②CVPR 2022 | Tsinghua open source DAT： Vision with deformable attention Transformer
③DAT Interpretation of the thesis ( warm wind )