unable to reproduce the results of BTCV, only a mean dice score of 0.2.

[XinYue-Yu1031]

Hello,

I trained BTCV on a single V100 GPU with a random data split ratio of 4:1, using a batch size of 1 and training for 3000 epochs. However, my mean dice score is only 0.2. Do you have any good training suggestions, or could you provide your BTCV training model? I would greatly appreciate it.

[ge-xing]

The training loss is very strange. I think you should try the BraTS dataset and see if there are still any problems.

[Tai-Yng]

Hello,can you tell me how is this dice and epoch graph given? Is the data plotted through the code again?Thank you!