Retinet

This project aims to train a CNN with the diabetiec retinopathy images as found on the kaggle competition on the subject.

We believe training a neural network on the train images and then using Visualization techniques like Deep Dream, we might be able to uncover hidden indicators of diagnostic and prognostic value in patients affected with Diabetic retinopathy.

Training a CNN

A large convolutional network is one option in achieving this, but it is very expensive

Classification

0 - No DR

1 - Mild

2 - Moderate

3 - Severe

4 - Proliferative DR