Overview
Skin Cancer Detection AI is a computer vision application that classifies types of skin cancer based on uploaded images. The system leverages a VGG-16 Convolutional Neural Network trained on thousands of pre-identified skin lesion images, delivering accurate classification directly in your browser.Quick start
Get up and running with the model in minutes
Demo
See the web interface in action
Key features
VGG-16 CNN architecture
Pre-trained convolutional neural network optimized for skin lesion classification
Browser-based inference
Real-time predictions using TensorFlow.js with no server required
7 classification categories
Identifies Actinic Keratoses, Basal Cell Carcinoma, Benign Keratoses, Dermatofibroma, Melanoma, Melanocytic Nevus, and Vascular Lesions
Simple web interface
Intuitive file upload with instant classification and confidence scores
Classification categories
The model can classify skin lesions into seven distinct categories:- Actinic Keratoses - Precancerous skin patches caused by sun exposure
- Basal Cell Carcinoma - Most common type of skin cancer
- Benign Keratoses - Non-cancerous skin growths
- Dermatofibroma - Common benign skin nodules
- Melanoma - Most serious type of skin cancer
- Melanocytic Nevus - Common moles or beauty marks
- Vascular Lesion - Abnormalities of blood vessels in the skin
This tool is designed for educational and research purposes. Always consult with a qualified healthcare professional for medical diagnosis and treatment.
How it works
The application uses a three-stage process:Image preprocessing
Uploaded images are resized to 75x100 pixels and converted to tensors for model input
Model inference
The VGG-16 CNN processes the image tensor and generates probability scores for each of the 7 categories
Technology stack
- Model: VGG-16 Convolutional Neural Network
- Training: Python with Keras and TensorFlow
- Inference: TensorFlow.js for browser-based predictions
- Frontend: HTML, JavaScript, and Materialize CSS
- Input size: 75x100x3 (height, width, RGB channels)