Skip to content

Deep Learning based Skin Cancer Detection using multiple CNN architectures (VGG, ResNet, DenseNet, EfficientNet, Inception) with image preprocessing using ESRGAN and performance comparison for clinical AI research.

Notifications You must be signed in to change notification settings

amitabh-7t/Skin-Cancer-Detection-using-Deep-Learning

Repository files navigation

🩺 Skin Cancer Detection using Deep Learning

An AI-powered skin lesion classification system leveraging advanced Convolutional Neural Networks to support early screening of skin cancer. Multiple models have been compared to identify the best architecture for accurate diagnosis.


🎯 Objective

  • Classify skin lesions using dermatoscopic images
  • Compare performance of multiple transfer-learning architectures
  • Support medical experts in early cancer risk detection

🧠 Deep Learning Models Used

Model Status Remarks
VGG-16 Transfer learning baseline
VGG-19 Improved feature extraction
ResNet50 Strong performance on medical images
DenseNet169 Captures deep hierarchical features
EfficientNetB0 Efficient & high accuracy
InceptionV3 Good at multi-scale feature learning
ESRGAN Used for resolution enhancement

📌 All training notebooks included inside the repository.


📂 Project Structure

📁 Skin-cancer
├─ content/                  # Image content / visual assets (optional)
├─ VGG-16.ipynb
├─ VGG-19.ipynb
├─ ResNet50.ipynb
├─ DenseNet169.ipynb
├─ InceptionV3.ipynb
├─ EfficientNetB0.ipynb
├─ ESRGAN.ipynb
└─ README.md

🧪 Dataset Used

Dermatoscopic skin lesion dataset:

  • HAM10000 / ISIC (commonly used in research)

(You can add dataset link here if public)


📊 Planned Evaluation Metrics

  • Accuracy
  • Precision, Recall, F1-Score
  • ROC-AUC Score
  • Confusion Matrix
  • Grad-CAM Visual Explanations 🔥

These results will help determine the best model for deployment.


🚀 How to Run

# Install dependencies (suggested)
pip install tensorflow keras numpy pandas matplotlib scikit-learn opencv-python
pip install jupyter
jupyter notebook

Execute any notebook like:

VGG-16.ipynb

🔮 Future Enhancements

  • Deploy as a web app (Streamlit or Gradio UI)
  • Add Explainability module (Grad-CAM heatmaps)
  • Clinical-level evaluation with dermatologists
  • Export to ONNX / TensorRT for real-time inference

⚠ Disclaimer

This project is for research & educational purposes only. It is not a replacement for professional medical diagnosis.


👨‍💻 Developer

Amitabh Thakur AI/ML Engineer | CSE (AI & ML) @ Dayananda Sagar University Founder — Humans Care Foundation Bangalore, India


🌟 Contributions, forks & research collaboration are welcome!

About

Deep Learning based Skin Cancer Detection using multiple CNN architectures (VGG, ResNet, DenseNet, EfficientNet, Inception) with image preprocessing using ESRGAN and performance comparison for clinical AI research.

Topics

Resources

Stars

Watchers

Forks