Adaptation of @titu1994's Inception v4 and Inception ResNet v4 architectures to MRI images of the human brain. The paper on these architectures is available at "Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning".
This repo serves as an example on how to run experiments on Google Cloud, not how to segment brain images. Today there are more efficient architectures out there for this kind of segmentation.
This repository contains code for training the networks to segment white matter and gray matter on MRI scans from the The Open Access Series of Imaging Studies (OASIS) archive.
To start the experiment, clone the repository and run
$ ./experiment.sh
Data is downloaded, extracted and preprocessed automatically.
Provision a Google Cloud CPU or GPU instance with google-cloud.sh using either of the following commands:
$ ./google-cloud.sh --create-cpu-instance
$ ./google-cloud.sh --create-gpu-instance
SSH into the instance once it is up and running, clone, and invoke experiment.sh from there.