Detection of Diabetic Retinopathy using CNNs

Problem Statemen

You’re a Computer Vision Engineer at health.ai. Your company is developing a deep learning application to automate the detection of diabetic retinopathy. The company is sourcing high-resolution retina image data from various clinical partners but the dataset is expected to be huge and cannot be stored on a central system. You’re asked to build a proof of concept using the Kaggle retinopathy dataset to train a CNN model with the Mirrored Strategy and deploy it with TensorFlow Serving.

Objective

To build a CNN model using distributed training that can detect diabetic retinopathy and deploy it using TensorFlow Serving.

Dataset Details

The dataset contains a large set of high-resolution retina images taken under a variety of imaging conditions. A left and right field is provided for every subject. Images are labeled with a subject id as well as either left or right. A clinician has rated the presence of diabetic retinopathy in each image on a scale of 0 to 4. Like any real-world dataset, you will encounter noise in both the images and labels. Images may contain artifacts, be out of focus, underexposed, or overexposed.

Link to the Dataset: https://www.dropbox.com/sh/7z7xq2lq3ogspcv/AACF_50dOtFaVYoII80abNPLa?dl=0

Prerequisites

TensorFlow

Keras

TensorFlow Serving

Solution

See the code - Jupyter Notebook

See Project Report

See entire project on GitHub