You can use encoder and decoder in http://sceneparsing.csail.mit.edu/model/pytorch/baseline-resnet50-upernet/ and edit opts as

You can use encoder and decoder in http://sceneparsing.csail.mit.edu/model/pytorch/baseline-resnet50-upernet/ and edit opts as
python -u test.py --model_path C:\Coding\Python\segmentation\baseline-resnet50_dilated8-ppm_bilinear_deepsup --test_img ADE_val_00001519.jpg --arch_encoder resnet50 --arch_decoder upernet --fc_dim 2048 --result C:\Coding\Python\segmentation\result

to run the test demo !

_Originally posted by @svjack in https://github.com/CSAILVision/semantic-segmentation-pytorch/issues/116#issuecomment-478235643_