John1231983

How about cardiac segmentation in kaggle challenge? you can refer VNet for detail. I am wondering if focal works better than crossentropy for semantic segmentation

@romainVala : I still wait the response from author

@romainVala: To run the code, you just download his caffe and compile it. After that, run training or just run develop (he already provide .caffemode) @TJKlein: I want to ask...

@YilinLiu97 : Not yet. I cannot run it code. I am still confusing number of classes. Do you know how to convert 207 classes in the dataset to 7 classes...

He used his caffe code. Check it README

@TJKlein : Could you please let me know how to map from 207 classes to 7 classes? Because the given dataset includes 207 classes.