✨ ECLIPSE – Offline Skin Lesion Classification

Fully Offline · Standalone MSI Installer · Parallel Swin Encoder . DenseNet-169

---

🚀 Overview

ECLIPSE is a standalone, fully offline skin lesion classification system built using a
Swin Transformer + DenseNet-169 + U-Net encoder architecture, deployed as a Windows MSI installer.

It runs entirely on-device, ensuring:

• 🔒 Complete data privacy
• ⚡ Fast processing
• 🌐 Zero internet dependency
• 🖥️ Seamless deployment on any Windows machine

---

🧠 System Architecture

🔄 Workflow

🏗️ Parallel Encoder — DenseNet-169 + U-Net

---

💻 Application Screenshots

🖼️ Input View

📟 Prediction Output

🔍 Classification UI

---

⭐ Features

• 🔌 100% Offline — No cloud, no external API calls
• 📦 Distributed as a Windows MSI Installer
• 🧠 ONNX Runtime for fast, local inference
• 🎯 Benign / Malignant classification with confidence %
• 📊 Optional CSV export
• 🔒 Guaranteed privacy — images never leave the device
• 🖥️ Clean and intuitive WPF interface

---

📥 Installation (MSI)

1. Download ECLIPSE v1.0.0.msi
2. Run the installer
3. Follow the setup wizard
4. Launch via:
   Start Menu → ECLIPSE – Skin Lesion Classifier

✔ No dependencies required
✔ No internet required
✔ Works instantly after installation

---

📘 Usage

1. Open ECLIPSE
2. Click Browse and select a dermoscopic image
3. Click Predict
4. View:
   • Benign / Malignant classification
   • Confidence score
5. (Optional) Export results as CSV

All inference happens locally using the embedded ONNX model.

---

📊 Results

📈 Confusion Matrix

---

📉 Training Curves

🔹 Training History

🔹 Fine-Tune Curves

---

👥 Contributors

Anagha P Kulkarni  
Debabrata Kuiry  
B Chiru Vaibhav  

---

Dataset used: Unified Dataset for Skin Cancer Classification on Kaggle

---

📄 License

Academic use only.

---