How to load the pretrained model ? about swin-transformer-semantic-segmentation HOT 6 CLOSED

Hi, @David-zaiwang . "upernet_swin_base_patch4_window7_512x512.pth" is a model finetuned on ADE20K, not an ImageNet pre-trained model.
You would better use pre-trained models held in this page

from swin-transformer-semantic-segmentation.

For people having the same issue~

Please try to use this command:

python tools/train.py \
  configs/swin/upernet_swin_base_patch4_window7_512x512_160k_ade20k.py \
  --options \
  model.pretrained=swin_base_patch4_window7_224.pth

The ImageNet pre-trained model swin_base_patch4_window7_224.pth can be downloaded from this table

from swin-transformer-semantic-segmentation.

For people having the same issue~

Please try to use this command:

python tools/train.py \
  configs/swin/upernet_swin_base_patch4_window7_512x512_160k_ade20k.py \
  --options \
  model.pretrained=swin_base_patch4_window7_224.pth

The ImageNet pre-trained model swin_base_patch4_window7_224.pth can be downloaded from this table

https://github.com/SwinTransformer/storage/releases/download/v1.0.0/swin_base_patch4_window7_224.pth

from swin-transformer-semantic-segmentation.

Hi @David-zaiwang what command did you exactly use then to fine tune on your custom dataset?
Thanks

from swin-transformer-semantic-segmentation.

Hi @David-zaiwang what command did you exactly use then to fine tune on your custom dataset? Thanks

Hello, have you solved this problem

from swin-transformer-semantic-segmentation.

@David-zaiwang @AnukritiSinghh @xuhao-anhe the same question!have you solved this question?
@impiga It seems that these pretrain offered by official swin cannot work well on this code?

from swin-transformer-semantic-segmentation.