search

FengheTan9 / Medical-Image-Segmentation-Benchmarks

A Pytorch implement of medical image segmentation U-shape architecture benchmarks
MIT License
72 stars 3 forks source link

Questions for CMUNeXt Training #2

Open WFLiu0327 opened 10 months ago

WFLiu0327 commented 10 months ago

Hello author, your article CMUNeXt is great. We have some questions about training with this repository and would like to get your answer:

  1. we divide the dataset and train CMUNeXt according to the readme document, however the best results obtained are much lower than the results in the paper as follows: epoch [212/300] train_loss : 0.1227, train_iou: 0.8474 - val_loss 0.2993 - val_iou 0.6664 - val_SE 0.7741 - val_PC 0.8161 - val_F1 0.7900 - val_ACC 0.9667 We believe this may be a problem due to inconsistent dataset partitioning. So if you can could you send me a list of the training set and test set if possible, thank you very much! Email: liuweifeng0327@gmail.com!!!
FengheTan9 commented 10 months ago

Thanks for using our code. Split file on BUSI I have sent to your email.

sainny commented 8 months ago

I also meetthe same problem, can you share the data set if possible,Email: liurunhuasainy@gmail.com