Diabetic Retinopathy and Glaucoma are the leading causes of blindness all around the globe. In countries like India there's drastic shortage of more than 100,000 eye doctors and just 6 million of the 72 million diabetics are screened for eye disease.
Both of My parents are ophthalmologist and they work at a organization called Aravind Eye Hospital. The organizations key vision and mission:
- Provide compassionate and quality eye care affordable to all
- Extending the reach of quality eye care to the poor and needy – through active community involvement, screening camps, and IT enabled Vision Centres in rural areas.
So I believe we can develop AI models to fill the gap and improve health care in rural India.
So the objective is to use pytorch to train two models
- Classification model for Diabetic Retinopathy diagnosis
- Segmentation model to compute cup to disc ratio for Glaucoma diagnosis
What it does
- One system for diagnosing both Diabetic retinopathy and Glaucoma.
- Predict diabetic retinopthy with very high accuracy
- Compute Cup to disc ratio with very accurate segmentation mask
How I built it
The is project is based on Pytorch
- I trained the diabatic retinopathy classification model using Fastai
- I trained the cup to disc segmentation model using pytorch -I built the backend using flask
Challenges I ran into
- Creating the Web App was the biggest challenge, even though it is a simple app I did not have much knowledge in web dev so it was challenging and I did not have a lot of time left.
Accomplishments that I'm proud of
- Creating a end to end application
What I learned
- I learnt how to use pytorch in detail (I generally used keras or fastai)
- I learnt a little bit of Webdev
What's next for EyeAI
- Improve the training accuracy of both the models and do a clinical validation.
Log in or sign up for Devpost to join the conversation.