This tutorial demonstrates how to use cross-modal data programming to train a Convolutional Neural Network on the OpenI chest radiograph dataset. The entire tutorial is contained within openi_demo/openi_demo.ipynb. For those who do not wish to execute the code, we have included openi_demo/openi_demo.html, which contains an HTML version of the executed notebook.
Steps to set up the tutorial are as follows:
(1) Install Snorkel MeTaL and requirements by running pip install --requirement python-package-requirements.txt
(2) Run download.sh within the openi_demo/data subdirectory
(3) Run jupyter notebook from the current directory
(4) Open the openi_demo/openi_demo.ipynb notebook
(5) Enjoy the tutorial!
Labeling functions (LFs) for each application studied in our recent work can be found in the lfs directory.