search

lucidrains / segformer-pytorch

Implementation of Segformer, Attention + MLP neural network for segmentation, in Pytorch
MIT License
342 stars 43 forks source link

The model configurations for all the SegFormer B0 ~ B5 #4

Open rose-jinyang opened 3 years ago

rose-jinyang commented 3 years ago

Hello How are you? Thanks for contributing to this project. Is the model configuration in README MiT-B0 correctly? That's because the total number of params for the model is 36M. Could u provide all the model configurations for SegFormer B0 ~ B5?

lucidrains commented 3 years ago

@rose-jinyang hmm, i'm actually not too sure, i did go back and reread the paper and noticed they used depth-wise convs in the feedforward https://github.com/lucidrains/segformer-pytorch/commit/5868d8e75b2360bf26a83bd03f2a8ec3d9b1e8f7 however, i don't think that's enough to account for the difference

rose-jinyang commented 3 years ago

I've updated the project to the new commit. But the number of paprams for MiT-B0 is 7M.

rose-jinyang commented 3 years ago

Hi @lucidrains Did u check the number of params of MiT-B0 model?

rose-jinyang commented 3 years ago

Hello @lucidrains Did u check the number of params of MiT-B0 model?

Sundragon1993 commented 3 years ago

Hi @rose-jinyang Have you found the solution?