DSC and accuracy

[Xtaotianxiawudi]

Dear author,
I retrained the ISIC2017 dataset using your code, but I found that the validation loss remains high and fails to converge. The DSC accuracy is only around 0.82, which is noticeably lower than the results reported in the paper. Could you please advise me on what I should do?

[wurenkai]

I found that you are currently only training for 150 epochs, and the screenshot you provided is the verification result at 150 epochs. Ultimately, we need to train to the last epoch and conduct tests using the best epoch. It is suggested that you can read the following train.py carefully to understand the specific training process. Furthermore, it is suggested that you read carefully the learning strategies we adopt to have a more comprehensive understanding of the convergence process.