search

Dootmaan / DSRL

Implementation of CVPR 2020 Dual Super-Resolution Learning for Semantic Segmentation
MIT License
95 stars 17 forks source link

some questions about datasets #5

Open lions-heart opened 3 years ago

lions-heart commented 3 years ago

tks for ur work. i have git clone ur code, and got some trouble on the datasets. e.g: while i run the main in 'cityscapes.py' it print the right picture, that image and label concatnate together. but it will return a distortion picture(colour went wrong) from dataloader in the 'train.py'. as a result the output_sr got a whole black picture and output_seg has the same problem.

Dootmaan commented 3 years ago

Hi there, thank you for your question. Dataloader in train.py will do augmentations for trainning datasets and that may be the reason of distortion. And for the whole black problem, please make sure the outputs are multiplied by 255.