jeya-maria-jose / UNeXt-pytorch

Official Pytorch Code base for "UNeXt: MLP-based Rapid Medical Image Segmentation Network", MICCAI 2022
https://jeya-maria-jose.github.io/UNext-web/
MIT License
459 stars 76 forks source link

The shifted MLP operation in the code seems to be different from what the paper says. Can you help me? #27

Closed yangliu1510 closed 1 year ago

yangliu1510 commented 1 year ago

From the code the shifted mlp operation seems a bit like the operation in [S2-MLP: Spatial-Shift MLP Architecture for Vision].

sdtegaotian commented 1 year ago

yes,I think so

chi-sq commented 1 year ago

yes,it totally different.I don't how shifted operation pays attention to the local as the paper illustrate. I wonder if my understanding is right.

jeya-maria-jose commented 1 year ago

The shift operation restricts the information that the MLP sees. Shifting across height and width simply localizes the span of information that the network sees.