Case Study: New York University builds a melanoma-detection app with Google for Education and Google Cloud Platform

Students build an app that uses skin imaging to identify moles as malignant or benign

New York University’s HackNYU 2018 team set out to build DermaScan, a mobile app to help users scan skin moles and identify whether they may be malignant or benign as an early step toward detecting melanoma. Working with Google for Education and Google Cloud Platform products including Compute Engine, App Engine, and TensorFlow, the students aimed to create a fast, accessible preliminary screening tool that could help people know when to see a dermatologist.

Google for Education helped the NYU team prototype a cross-platform Android/iOS app and retrain Google’s Inception image classification model on their own skin-mole dataset using GPU-enabled Compute Engine instances, which significantly sped up model training. The result was a working machine-learning prototype backed by REST APIs and a Python Flask backend, with the team continuing to improve accuracy, reduce false negatives, and expand the app’s potential for remote consultation and broader skin-condition detection.

New York University

Harshit Srivastava

New York University