| 🚀 System Boot Loading | 🏠 Landing Hero |
|---|---|
 |
 |
| ⚙️ Clinical AI Pipeline | 👥 Role Portal Selection |
|---|---|
 |
 |
| 🩺 Patient Diagnostic Portal | 🏥 Clinical Doctor Dashboard |
|---|---|
 |
 |
CutisAI is a state-of-the-art deep learning system for automated skin lesion segmentation, classification, and risk assessment using dermoscopic imagery.
Trained on over 33,000+ ISIC images, CutisAI provides real-time clinical triage assistance in under 2 seconds on standard CPU hardware using INT8-quantized ONNX models.
- 🎯 Lesion Boundary Segmentation: High-precision pixel mask extraction using ResUNet.
- 🏷️ Multi-Class & Malignancy Classification: Rapid risk profiling powered by EfficientNet-B0.
- ⏱️ Sub-2-Second CPU Inference: Quantized INT8 ONNX execution for instant results.
- 📄 Clinical PDF Report Generator: Export structured diagnostic summaries with mask overlays and clinical recommendations.
- 👥 Dual-Role Interface: Dedicated portals tailored for both Patients and Clinicians.
- 🧠 Two-Stage AI Pipeline: Sequential segmentation and classification for robust ROI analysis.
- 🎨 Modern Dark Aesthetic: Custom Glassmorphism UI with responsive CSS design tokens.
- ⚡ Optimized Inference: INT8 ONNX Runtime backend eliminating heavy GPU dependencies.
- 📊 Interactive Analytics: Confidence score visualization and diagnostic breakdown.
- 🩺 Clinical Recommendation Engine: Automated risk category triage (Low, Moderate, High).
- 📱 Fully Responsive: Optimized for desktop monitors, clinical tablets, and mobile devices.
graph TD
A[Dermoscopic Image Input] --> B[FastAPI Inference API]
B --> C[Preprocessing & Normalization]
C --> D[ResUNet Segmentation Model]
D --> E[Binary Lesion Mask Extraction]
C --> F[ROI Crop & Mask Fusion]
F --> G[EfficientNet-B0 Classifier]
G --> H[Risk Score & Probability Vector]
H --> I[React Frontend Dashboard / PDF Generator]
| Domain | Tech / Framework | Description |
|---|---|---|
| Frontend | React 18, Vite 5, TypeScript | SPA client with Vanilla CSS tokens & Glassmorphism |
| Backend | Python 3.10+, FastAPI | High-performance async REST API endpoint |
| Machine Learning | PyTorch, ONNX Runtime | ResUNet + EfficientNet-B0 (INT8 Quantized) |
| PDF Engine | HTML2Canvas / jsPDF | Dynamic clinical PDF report generation |
| Dataset | ISIC 2018 / 2019 / 2020 | 33,000+ annotated dermoscopic lesion images |
| Metric | Result | Target Benchmark |
|---|---|---|
| Dice Score (DSC) | 0.9007 | Segmentation Accuracy |
| AUC-ROC | 95.6% | Classification Discrimination |
| Inference Latency | < 2.0 seconds | CPU execution (INT8 ONNX) |
| Dataset Size | 33,000+ | ISIC Dermoscopic Images |
CutisAI/
├── frontend/ # React SPA (Vite + TypeScript)
│ ├── src/
│ │ ├── components/ # Topbar, LoadingScreen, ResultPanel, Buttons
│ │ ├── pages/ # Landing, UserHome, UserResult, DoctorDashboard
│ │ ├── utils/ # generatePdf.ts
│ │ └── routing/ # AppRouter & lazy routes
├── backend/ # FastAPI Python service
│ ├── src/ # API routes & ONNX inference runner
│ └── models/ # ONNX INT8 model weights
├── training/ # PyTorch model training & evaluation scripts
├── docs/ # PRD, TechStack, & Design System docs
├── images/ # Readme preview assets & screenshots
└── README.md
git clone https://github.com/yashraj10messi/Cutis-AI.git
cd Cutis-AIcd frontend
npm install
npm run devFrontend will start on
http://localhost:3000(orhttp://localhost:3001).
cd backend
pip install -r requirements.txt
uvicorn src.main:app --reload --port 8000Backend API docs available at
http://localhost:8000/docs.
Important
CutisAI is a research and educational prototype and is not a certified medical device. Predictions generated by this system are strictly advisory and should never replace evaluation, diagnosis, or treatment by a licensed dermatologist or medical professional.
- Author: Yash Raj Sharan
- Contact: yashraj10messi@gmail.com
- License: MIT License — Copyright (c) 2026 Yash Raj Sharan
⭐ If you find CutisAI helpful, consider giving it a star on GitHub!