search

YichiZhang98 / DTML

[PRCV 2021] The official code for "Dual-Task Mutual Learning for Semi-Supervised Medical Image Segmentation"
24 stars 2 forks source link

how to get best model #5

Closed 18677064404 closed 3 years ago

18677064404 commented 3 years ago

Dear Yichi Zhang：

Hi, Thanks for your great work! I am intersted in your work.

Sorry to disturb you.I have some qustions. I want to know how to get the best model files : ". /model/DTML_16labels/best_model_dis.pth" and ". /model/DTML_16labels/best_model_seg.pth".

Can you tell me how to do it? Thanks

Best wish

YichiZhang98 commented 3 years ago

Hi, thanks for your interest. The trained best model files can be found at https://github.com/YichiZhang98/DTML/tree/main/model/DTML_16labels Best regards

18677064404 commented 3 years ago

Thanks for your reply.Yeah, I see the best model files are already in the model directory.  When I try to adjust parameters to generate a new model directory, I must put the "bestmodel.pth" in model directory first and then run the test_LA.py. So I want to know how to get the two "bestmodel.pth" files by training. If there need to add code to generate the two files. Thank you again!

If you are an Chinese, you can read the following: Yichi Zhang 先生您好！       首先，感谢您的回复。       是的，我在model文件目录下看到了这两个"bestmodel*.pth" 文件。但是当我想要调整参数，生成一个新的model文件时，训练完后，需将这两个最佳模型文件放置在这个model文件夹下，才能进行测试。       我想知道，如何通过训练得到这两个最佳模型文件呢？是不是需要增加代码去生成它呀？       再次感谢您！

------------------ 原始邮件 ------------------ 发件人: @.>; 发送时间: 2021年7月6日(星期二) 中午12:37 收件人: @.>; 抄送: @.>; @.>; 主题: Re: [YichiZhang98/DTML] how to get best model (#5)

Hi, thanks for your interest. The trained best model files can be found at https://github.com/YichiZhang98/DTML/tree/main/model/DTML_16labels Best regards

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.

YichiZhang98 commented 3 years ago

Hi, the trained models are initially generated in "model/LA/DTML_settings/". Here, we use the final checkpoint (iter6000*.pth) as the best model for testing.

18677064404 commented 3 years ago

Hi,Thank you very much. Your reply has helped me a lot! Wish you the best of luck in your work!

------------------ 原始邮件 ------------------ 发件人: "YichiZhang98/DTML" @.>; 发送时间: 2021年7月7日(星期三) 上午10:05 @.>; @.**@.>; 主题: Re: [YichiZhang98/DTML] how to get best model (#5)

Hi, the trained models are initially generated in "model/LA/DTML_settings/". Here, we use the final checkpoint (iter6000*.pth) as the best model for testing.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.

18677064404 commented 3 years ago

Hi,I'm sorry for disturbing you again.

I run the file with "iter_6000_seg.pth", the result showed:average metric is [ 0.18605297 0.10368403 53.06831007 23.19028863]. I try your best model file got the result as your paper showed.

Before running the "test_LA.py", I downloaded the "2018 Atrial Segmentation Challenge COMPLETE DATASET",and then used the "la_heart_processing.py" file to process the "Training Set", and put in in the corresponding directory，and run the file “train_la_dtml” , training success.

So can you give some advice, or is there something I didn't deal with? Thanks.

YichiZhang98 commented 3 years ago

Hi, I have not never been in a situation like yours. Maybe you can check if there is something wrong with the training set or testing set. Best regards