README.md

metadata

title: DR Classification
emoji: 🐨
colorFrom: gray
colorTo: blue
sdk: streamlit
sdk_version: 1.44.1
app_file: app.py
pinned: false
license: mit

🐨 DR Classification

This is a Streamlit-based web app for Diabetic Retinopathy (DR) Classification using fundus images. The model classifies retinal images into different DR severity levels to assist in early detection and monitoring.

💡 Features

• Upload a fundus image and get an instant DR classification.
• Preprocessing pipeline (CLAHE, gamma correction, normalization, etc.) to enhance input quality.
• Uses a fine-tuned DenseNet-121 model pretrained on ImageNet.
• Supports visual output like prediction label and optionally Grad-CAM heatmaps for model explainability.

🖼 Dataset

The dataset used is uploaded on the Hugging Face Hub:
👉 your-username/your-dataset-name

It includes fundus images categorized into the following DR stages:

• 0: No DR
• 1: Mild
• 2: Moderate
• 3: Severe
• 4: Proliferative DR

🚀 How to Use

1. Click the “Open in Spaces” button or visit the live app.
2. Upload a fundus image (JPEG or PNG).
3. View the model prediction and (optional) heatmap.

🧠 Model Details

• Architecture: DenseNet-121
• Pretrained on: ImageNet
• Fine-tuned on: Fundus images from the uploaded dataset

🛠 Tools & Libraries

• Streamlit
• PyTorch / TensorFlow (depending on what you're using)
• OpenCV for image preprocessing
• Hugging Face Datasets

📄 License

This project is licensed under the MIT License.

---

Check the app 👉 Live Demo