Suggestion: replace dropout with batch norm

MrGiovanni / UNetPlusPlus

[IEEE TMI] Official Implementation for UNet++

Other

2.26k stars 538 forks source link

Suggestion: replace dropout with batch norm #2

Closed tybrad11 closed 5 years ago

tybrad11 commented 6 years ago

I'm testing a small data set (training on 10 CT liver segmentations, testing on 4), but I get WAY better results when I replace dropout with batch norm in the standard_unit(). May be data size dependent, but worth noting.

wl202305 commented 5 years ago

hello，now,i am look this project.because i am a beginner.can you tell me the 'main' file of this project?thank you very much.my email:wl202305@126.com

MrGiovanni commented 5 years ago

@tybrad11 : Thanks for your suggestions! We have updated the backbones to various famous architectures in computer vision (see our README).

MrGiovanni commented 5 years ago

@wl202305 : We have updated the README. You can use 'BRATS2013_application.py' to run the code, also, you can use our code examples for your own data. Thanks!