Overview

A CNN that reads in ultrasound images and manually labeled masks delineating nerve regions and reconstructs nerve masks for test ultraound images.

Dependencies

• Python 2.7+ (pip install python2.7)
• opencv/cv2 (install through Homebrew -- Mac installation follow instructions here: https://www.learnopencv.com/install-opencv-3-on-yosemite-osx-10-10-x/)
• keras (pip install keras)
• Theano (pip install Theano)
• sklearn (pip install sklearn)
• Numpy (pip install numpy)

Basic Usage

Download data from https://www.kaggle.com/c/ultrasound-nerve-segmentation/data (File Size ~ 1 GB)
Samples of test and train data present in folder

root directory should look like:

../P5_Submission_Folder
-raw
|
---- trainsample
| |
| ---- 1_1.tif
| |
| ---- …
|
---- testsample
|
---- 1.tif
|
---- …

Change file paths in capstone.py that are marked by comments
Run python capstone.py from terminal