search

lhoyer / DAFormer

[CVPR22] Official Implementation of DAFormer: Improving Network Architectures and Training Strategies for Domain-Adaptive Semantic Segmentation
Other
470 stars 93 forks source link

What is the config to run the baseline without any extra method? #28

Closed xingchenzhao closed 2 years ago

xingchenzhao commented 2 years ago

Thank you for the great work! I wonder what is the config to run the baseline using pure transformer without the proposed method?

lhoyer commented 2 years ago

Thanks for your interest in our work! All configs used in the paper can be generated using python run_experiments.py --exp <ID>. As you seem to refer to table 1, the relevant ID is 1. You can find more details here: https://github.com/lhoyer/DAFormer/blob/master/experiments.py#L223 In order to only train the transformer, you can comment out all architectures except 'segformer' in models. To only run UDA training, you can comment out all variants except 'dacs' in udas.

xingchenzhao commented 2 years ago

Thank you so much for your quick response!! I will try your solution soon.