Skip to content

woodywff/nas_3d_unet

Repository files navigation

NAS-3D-U-Net

Introduction

This is the source code of this article: Neural Architecture Search for Gliomas Segmentation on Multimodal Magnetic Resonance Imaging

Main.ipynb is the main storyline which is self-explained. (If you don't wanna run it in Jupyter Notebook, don't forget to change the tqdm.notebook to tqdm.)

Most of the configurations are set up in config.yml.

I've put the three downloaded and unzipped dataset files in a nas_3d_unet/data/ folder as

nas_3d_unet/data/MICCAI_BraTS_2019_Data_Training/
nas_3d_unet/data/MICCAI_BraTS_2019_Data_Validation/
nas_3d_unet/data/MICCAI_BraTS_2019_Data_Testing/

You are free to keep these things anywhere else, just don't forget to change the corresponding arguments in the config.yml.

This repository is also an update for the previous brats2019 pipeline.

Development Environment

CUDA10 torch==1.2.0 torchvision==0.4.0

GTX1060 (6GB GPU Memory) is good enough for running the whole project (both searching and training) with patchsize=64.

GTX1080Ti (11GB GPU Memory) is recommended.

Acknowledgment

About

Neural Architecture Search for Gliomas Segmentation on Multimodal Magnetic Resonance Imaging

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published