Diabetic-Retinopathy-Prediction

A Neural Network built on Keras using Sequential Model to predict whether a person has Diabetic Retinopathy or not.

Model : Sequential

Layers : 4

Activation Layout : relu->relu->sigmoid

Output layer has been set to sigmoid to get uniform outputs of 0 or 1

Optimizer : adam

This model can be used for binary classification with dataset containing no string values.

Train Accuracy : 85%

Test Accuracy : 76%

Dataset used : Diabetic Retinopathy Debrecen Data Set

Next Steps : Implement a model to increase accuracy and reduce bias.

---

Environment Setup

Use the following commands for installing the libraries

$ pip3 install numpy

$ pip3 install pandas

$ pip3 install keras

---

How to Use

The dataset file should be in the same directory as the model

Command $ python model.py

The following files will be generated upon successful execution of the model model.h5 model.json

Update (Jan 2019)

Major refactoring and a different architecture coming up.