Ocumeda Machine Learning Coding Challenge

The task of this coding challenge is to develop a simple baseline approach for glaucoma classification on retinal fundus images.

Goals

1. Implement a machine learning approach for classification of glaucoma on retinal fundus images. Your approach should make sense as a simple baseline method.
2. Evaluate your approach in a meaningful way that would allow a fair comparison to other methods.
   • Define 1-3 meaningful metrics
   • Develop a meaningful experimental setup in terms of training/test/validation split and cross validation
3. Present your approach and key results (10 minutes max). Please keep in mind that the target audience has diverse educational backgrounds.

Data

We are using the data from the AIROGS Glaucoma Screening Challenge. You can downnload the data from the challenge website.

Implementation

You should implement your approach in Python 3. For your ML approach, you can use libraries such as torch, tensorflow, keras, scikit-learn, scikit-image, pandas, pytorch-lightning. 💡Hint: As a starting point, you might find the discussions from the Kaggle Diabetic Retinopathy Challenge useful.

Next Steps and Interview

Please implement your approach in a private GitHub fork of this respository and send us the your code at least 12 hours before your 2nd interview. Please structure the repository such that we can reproduce a run (without data paths). For providing us the code, please send us a zip-export of the forked repo.