****************** MUST SEE **********************

Before using app please enable the permission:

Settings >> Apps >> Diabetic Retinopathy >> Permissions >> Storage (Enable)


The application is powered by Tensorflow Deep Neural Network architecture. Model is trained and tested on 16,798 fundus images. Final accuracy of the model is 73.3 %.

Application Features :

1. Instant Results

2. Does not require an internet connection to run.

3. Neural network based image classification. 28 layer deep network.

4. Model is trained on 16,798 fundus images.

5. Can be used as a pre-screening tool for Diabetic Retinopathy.

6. Classifies image into two categories: Diabetic Retinopathy or No Diabetic Retinopathy.

Steps to use the application:

1. First, press the "load_image" button and upload an image from your gallery.

2. Once the image is uploaded then press "process" button to see the result.

The future update will include more categories for classification: Moderate, proliferative and severe Diabetic Retinopathy

Suggestion :

If one wants to take fundus image using the cell phone camera then one can attach a zoom lens to the cell phone and then take the picture of a dilated eye.

For more information visit:

www.shoravsuriyal.com