-
Notifications
You must be signed in to change notification settings - Fork 72
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
results #13
Comments
Hi jie, I appreciate your interest in our work. I would like to ask that you limit your issues to one thread because you have re-opened issues that are not related to code reproducibility is in #11 and #10. Can you try to reproduce the ISIC results? I have recently moved to pytorch and have reproduced ISIC without any struggles. |
Thank you for your reply!First of all, you have processed the data to 128 * 128 in the bus code, but there is no relevant processing for the ISIC code, so I think it is easier to reproduce the bus code than isic's. Secondly, I think the cross validation should not be too different, but I will review the code well according to your ideas. In addition, is it convenient for you to share your Pytorch code? Maybe I'm a little more familiar with Pytorch. It doesn't matter if it's inconvenient. Thank you. |
Hi, I try to use my data with this code, but why I get the dsc>1. Like this: |
Hi Kevin, can you check the range of your predictions and the range of your ground truth? Both masks should be in range [0-1]. You can just print the max and check what it is. Also, try adjusting the alpha and gamma parameters. Maybe your dataset is easier to segment and so the network is becoming super biased to prediction TPs and FPs. |
Hey @nabsabraham, any chance that you can share the pytorch version? I really would like to try it |
Hi @luistelmocosta, thanks for your interest! Is it the model you were looking for or just the loss function? If it is the latter, I just wrote this up quickly but I think it should work:
If it's the model, I believe you can get started with oktay's pytorch version of attention networks. I will have to rewrite it because I never hung on to it but its essentially a few extra layers added on to oktay's model. |
Thank you for the quick reply. I am currently using this version:
Does this look good or should I had Thank you |
Hi there,
Thanks for sharing your code. I try to reproduce your article and use your code intact, but the experimental results are quite different from the experimental results in your paper. I would like to ask you what you need to pay attention to in the process of reproducing the code.
model:attn_reg,loss:focal_tversky
my results:
DSC 0.748
Precision 0.860
Recall 0.752
results in article:
DSC 0.804
Precision 0.829
Recall 0.817
Hope to get your help! thanks!
The text was updated successfully, but these errors were encountered: