Leonardo Kulon
← Projects

Skin Cancer Detector

A CNN that classifies skin lesions as benign or malignant from a live webcam feed, with 86% accuracy.

Role
Solo developer
Date
Links
GitHub

How it works

  • Keras model fine-tuned from VGG16 on the HAM10000 dermatoscopic image dataset
  • Several training iterations with data augmentation, reaching 86% accuracy
  • OpenCV script classifies webcam frames in real time