search

lhoyer / HRDA

[ECCV22] Official Implementation of HRDA: Context-Aware High-Resolution Domain-Adaptive Semantic Segmentation
Other
233 stars 31 forks source link

Binary instance segmentation masks and domain generalization with unlabeled real data #44

Open LuanSilvaTec opened 6 months ago

LuanSilvaTec commented 6 months ago

Thank you very much for sharing the code, excellent work. I'm trying to train the framework using synthetic data for domain generalization with unlabeled real data. Is there any documentation regarding this? When using my own data (synthetic and real), both with a binary segmentation mask (grayscale), I am having problems with the IoU and Acc metrics, which have the value "Nan", the only class with values (above 90% inclusive) is background. Any post-processing recommendations for the labels? Thank you so much, this is an amazing work!

lhoyer commented 3 months ago

Hi @LuanSilvaTec,

Thank you for your interest in our work! You can find the instructions to use HRDA in a domain generalization setting here: https://github.com/lhoyer/HRDA?tab=readme-ov-file#domain-generalization

Best, Lukas