This project uses a Convolutional Neural Network (ResNet-18) model to detect common retinal diseases from fundus images. It supports detection for the following categories:
- 🟢 Normal
- 🔴 Cataract
- 🟠 Diabetic Retinopathy
- 🔵 Glaucoma
The frontend is built using React and hosted on Vercel, while the backend model is served via FastAPI and Hugging Face Spaces.
This application assists in the early detection of retinal diseases using AI, which can be critical in preventing vision loss. Given an image of a retina, the model classifies it into one of the predefined disease categories.
- Ophthalmologists
- Healthcare screening assistants
- Medical students and researchers
| Layer | Technology |
|---|---|
| Frontend | React + Tailwind CSS |
| Backend | FastAPI (Python) |
| Model | PyTorch (ResNet-18) |
| Hosting | Vercel (Frontend) |
| Model API | Hugging Face Spaces |
- Architecture: ResNet-18
- Input: Fundus image (resized and normalized)
- Output Classes: Normal, Cataract, Diabetic Retinopathy, Glaucoma
- Training Dataset: Public dataset from Kaggle (Retinal Eye Disease)
git clone https://github.com/VJnCode/Eye-Disease-Detection.git
cd Eye-Disease-Detection
⚠️ Note:The main backend API file is hosted on Hugging Face Spaces and cannot be opened directly like a code file. To view or test the API, navigate to 👉 [https://huggingface.co/spaces/NAVARASA/eye-disease-prediction]