Skip to content
A classifier written in native Keras that can be used to identify seven different types of skin lesions.
Python
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
tfjs-models Add files via upload Dec 13, 2018
HAM10000_metadata.csv Add files via upload Dec 10, 2018
README.md Update README.md Dec 13, 2018
model.py
model.zip Add files via upload Dec 10, 2018
preprocessing.py Add files via upload Dec 10, 2018

README.md

Skin-Lesion-Classifier

A classifier written in native Keras and converted to TensorFlow.js that can be used to classify seven different types of skin lesions.

This model uses the MobileNet architecture and was retrained on the HAM10000 dataset which contains approximately 10,000 documented images of skin lesions. https://arxiv.org/abs/1803.10417

The goal of the project was to deploy the final model as a web application. The live version can be found at: https://alexyu.ca/projects/Skin-Lesion-Classifier/

The model obtains an accuracy of approximately 80%, with a top-3 accuracy of approximately 94%.

You can’t perform that action at this time.