search

lucidrains / segformer-pytorch

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

Models weights + model output HxW #5

Open isega24 opened 3 years ago

isega24 commented 3 years ago

Hi,

Could you please add the models weights so we can start training from them?

Also, why you choose to train models with an output of size (H/4,W/4) and not the original (HxW) size?

Great job for the paper, very interesting :)

isega24 commented 3 years ago

Hi,

Hi,

Could you please add the models weights so we can start training from them?

With this I meant trained on the datasets of your paper so the testing would be faster.

Also, why you choose to train models with an output of size (H/4,W/4) and not the original (HxW) size?

I needed this because on my problem the resolution of other models are the same (H,W). I think upscaling the output of Segmenter or downscaling the other models outputs is not fair.

Great job for the paper, very interesting :)

QIU023 commented 3 years ago

I guess that checkpoint (pretrained weight) you need to obtain from NVlab, and also the keys in the state_dict are not corresponding to this one since the module names are different.