question about train uncertainty rectified pyramid consistency

[wener-yung]

Thanks for the excellent job！

I have some problems with the train-uncertainty-rectified-pyramid consistency experiment, and the experimental results are poor when I transform the code to my datasets. It is very insteresting to realize semi-supervised segmentation using a single model, but one single model cannot correct the training errors for the unlabeled data. I have upload some intermediate predictions for the unlabeled training data from epoch 60 and epoch 75. It is observed that p0 is not good and it tends to learning the serrated edges because of the low resolution of the outputs_aux1, outputs_aux2 and outputs_aux3.

It would be very grateful if you could clear up my confusion.
imgs.zip

[Xiangde Luo]

Hi,
Firstly, I did not evaluate the URPC on the Polyp Segmentation task, so it's hard to analyze these phenomena, in my opinion, you can try just two different predictions in URPC firstly and compare them with other semi-supervised methods. Then, for semi-supervised medical image segmentation tasks, the different datasets should need to finetune the consistency designs. In addition, we provide the ACDC and BraTS training and testing scripts, you can try them.
Best,
Xiangde.