crmo-diagnosis-using-mri

Setup

1. Install Miniconda

• Conda is a package manager that sandboxes your project’s dependencies in a virtual environment
• Miniconda contains Conda and its dependencies with no extra packages by default (as opposed to Anaconda, which installs some extra packages)

2. cd into src, run conda env create -f environment.yml

• This creates a Conda environment called cs229-project

3. Run source activate cs229-project

• This activates the cs229-project environment
• Do this each time you want to write/test your code

Data

• data.csv: list of image pairings and class curated from radiologist information
• legs_folder: excluded from github for privacy, but expected to contain MRIs.

Baseline

Initial experimentation with histograms is contained in Baseline.ipynb

Inception (folder /inception)

Jupyter notebooks as well as pickled output data and generated models are found in the inception folder. Key files:

• Tensorflow Inception v3 extraction.ipynb: contains pre-processing code for input images, and feature generation through inception model.
• inception_cnn_features2.pkl: selected set of features for images generated from tensorflow (dissimilarity of feature vectors, augmented, normalized and feature reduced)
• Inception CNN retraining.ipynb: various models trained on data generated with tensorflow.

Ensemble

• Ensemble Voter.ipynb: contains various models produced by voting ensembling inception and vbow models.

Visual bag of words

Work is contained in the vbow directory, which contains serialized models and intermediate data. Research work is spread through three notebooks: zach-cv.ipynb, zach-bagging.ipynb, and zach-models.ipynb. src/vision.py contains most of the core functionality, while the notebooks stitch things together.

Miscellaneous

• src/generate_features.py: contains utilities for feature generation, working with labels, and data augmentation.
• src/utils.py contains utilies for graphing, metrics evaluation, and analysis.