Ke Sun
Ke Sun
Please check the directory of data.
When you testing, you can set "sv_pred=True". https://github.com/HRNet/HRNet-Semantic-Segmentation/blob/8e4c0c3076ff319dcc8354a971b29c1250c16b1d/lib/core/function.py#L154
Because this model is trained based on pytorch-0.4.1, you should also run the test based on pytorch-0.4.1. The BN is different between pytorch-0.4.1 and pytorch-1.1, which results in worse performance.
Could you provide your testing settings and test the model on the val firstly?
Yes. Due to the limitation of submission, you can test the model on the val. Then, we can check the problem together.
@MandyMo I have tested the model based on Pyotrch-0.4.1. You should get MIoU 91.97 on val.
@Wangzhuoying0716 Sorry for late reply. I checked your log. When training on cityscapes, we use the "class balance" in the loss function. You can change the config and set the...
@whiteinblue Class weights are used to train the model on Cityscapes. If you don't want to use it, you can change it in the yaml.
@whiteinblue Class_balance is used in the loss function.
The segmentation mask is a one-hot label map. But when using softmax-loss, you need generate the predictions with the num_classes channels.