AI-powered skin lesion classification using ResNet-50, trained on the HAM10000 dataset. Upload a photo, get an instant prediction.
Deployed on Hugging Face Spaces
Four steps from photo to prediction
Upload a dermoscopy or regular photo of a skin lesion via the web interface
Hit the Analyze button — the image is sent to the FastAPI inference service
ResNet-50 classifies the lesion across 7 categories with a confidence score
See the predicted class, risk level, confidence chart, and all probabilities
ResNet-50 fine-tuned on HAM10000
7 lesion types across malignant, pre-cancerous, and benign categories
Most dangerous skin cancer, requires immediate attention
Most common skin cancer, highly treatable if caught early
Rough, scaly patch that can develop into skin cancer
Common moles, usually harmless
Non-cancerous skin growths
Harmless fibrous nodules
Blood vessel abnormalities on the skin
Built end-to-end — from training to deployment