Live App
Demo Video
Figma Prototype

Diabetes is one of the most prevalent chronic disorders, affecting millions in the U.S. alone. Poorly managed blood sugar can damage blood vessels in the eyes, potentially leading to blindness—a condition known as diabetic retinopathy.

Early detection is key. VisinEx was built to provide low cost, rapid, accessible screening for diabetic retinopathy using machine learning and an intuitive user interface.

Diabetic retinopathy often goes undiagnosed until it's too late. Access to ophthalmological exams can be limited or expensive. We aimed to create a tool that could empower individuals to screen themselves using just their phone.

VisinEx enables users to:

• Upload or capture a retinal fundus image
• Get predictions of diabetic retinopathy severity (None, Mild, Moderate, Severe, Proliferative)
• Receive AI-generated explanations and management advice in patient-friendly language

We divided our project into three key components:

1. ML Model: ResNet18-based image classifier to predict retinopathy severity
2. Gemini API: Generates simple explanations and condition management advice
3. Streamlit App: Provides an easy-to-use interface for users

We designed a custom-built smartphone fundus camera attachment for smartphones used for retinal imaging:

• Python
• PyTorch – ML model (ResNet18 + Transfer Learning)
• Gemini – AI-generated health explanations
• Streamlit – Web app deployment
• Figma – UI/UX design

Made with 💙 by the VisinEx team at HackDavis 2025:

• Toniya Patil - @tbpatil
• Joselyn Romero - @Joselyn15
• Abrar Sadikeen - @AbrarSadikeen