Closed qjl988 closed 8 months ago
Please reference/medical_seg/mmsegmentation/configs/pspnet/pspnet_r50 - d8_4xb4-40 k_voc12aug - 512 x512. Set in py. As far as I know, you first need to convert loss_decode=dict(type='CrossEntropyLoss', use_sigmoid=True, loss_weight=1.0)), use_sigmoid is set to False.
Please reference/medical_seg/mmsegmentation/configs/pspnet/pspnet_r50 - d8_4xb4-40 k_voc12aug - 512 x512. Set in py. As far as I know, you first need to convert loss_decode=dict(type='CrossEntropyLoss', use_sigmoid=True, loss_weight=1.0)), use_sigmoid is set to False.
Thank you very much. After setting use_sigmoid to False, it have solved.
When I was replicating the code on the PASCAL VOC dataset, I encountered the following error after changing num_classes to 21. I really don't know how to solve it and hope to get some help.