DermaAI

AI-Powered Skin Lesion Detection & Classification System

---

Overview

DermaAI is a deep learning-based web application that performs automated skin lesion segmentation and classification using dermoscopic images.

The system helps in early detection of melanoma by providing:

• Fast predictions
• Lightweight deployment
• Clinically meaningful insights

Users can upload an image and receive:

• Segmentation mask
• Classification (Benign / Malignant)
• Confidence score
• Clinical recommendation

---

Features

• 🧠 Two-stage AI pipeline (Segmentation + Classification)
• 🖼️ Lesion segmentation using U-Net / ResUNet
• 🔍 Classification using EfficientNet-B0 / ResNet-18
• ⚡ Fast inference (< 3 seconds on CPU)
• 🌐 Full-stack web app (React + FastAPI)
• 📊 Confidence score with probability visualization
• 🩺 Clinical recommendation system
• 📄 Downloadable report (PDF)

---

Architecture

User Image → Preprocessing → Segmentation Model (U-Net)
           → Masked ROI → Classification Model → Prediction
           → Result Dashboard (Frontend)

---

Tech Stack

Frontend:

• React.js
• Tailwind CSS

Backend:

• FastAPI

Machine Learning:

• PyTorch (Training)
• ONNX Runtime (Inference)

Deployment:

• Docker
• Nginx

Dataset:

• ISIC 2018 / 2019 / 2020

---

Model Details

Component	Model Used
Segmentation	ResUNet (U-Net + ResNet-18)
Classification	EfficientNet-B0 / ResNet-18
Loss Function	Dice Loss + Focal Loss
Optimization	ONNX Export + INT8 Quantization

---

Performance Metrics

• Dice Score: ≥ 0.85
• IoU: ≥ 0.78
• AUC-ROC: ≥ 0.88
• Inference Time: < 3 seconds (CPU)

---

Project Structure

DermaAI/
├── frontend/          # React frontend
├── backend/           # FastAPI backend
├── models/            # Trained models (ONNX / PyTorch)
├── data/              # Dataset (not included in repo)
├── notebooks/         # Training & experimentation
├── requirements.txt
└── README.md

---

▶ How to Run

1. Clone the Repository

git clone https://github.com/your-username/DermaAI.git
cd DermaAI

2. Backend Setup

cd backend
pip install -r requirements.txt
uvicorn app:app --reload

3. Frontend Setup

cd frontend
npm install
npm start

---

Use Cases

• 👨‍⚕️ Dermatologists – AI-assisted triage tool
• 🎓 Students – Learning medical AI systems
• 🔬 Researchers – Model experimentation & benchmarking

---

Disclaimer

This project is a research prototype and not a certified medical device. Predictions are advisory only and should not replace professional medical diagnosis.

---

Future Improvements

• Multi-class classification (beyond benign/malignant)
• Transformer-based models (SegFormer, Swin-UNet)
• Mobile deployment (on-device inference)
• Real-time camera input
• Clinical compliance (HIPAA/DISHA)

---

Credits

• ISIC (International Skin Imaging Collaboration) Dataset
• PyTorch
• FastAPI
• ONNX Runtime

---

⭐ Support

If you like this project, consider giving it a ⭐ on GitHub!