This project was built for The University of Texas at Austin’s “Elements of Mobile Computing” course. It was also a way for me to get started playing with Tensorflow by retraining the Tensorflow Inception-v3 network. However, there were many flaws, which I’ll touch on in the “What’s Next” section
Melanie was trained on several thousand images of moles, each classified as a mole, misshapen mole, or melanoma. The full ecosystem involves an iOS application, a web server, and a classifier. The mobile application allows you to take a picture of a mole, which it then processes and sends to the web server, which runs it through the classifier and returns a diagnosis.
Looking back, there’s a lot of things I wish I’d done differently. I’d re-write the web server to use Flask instead of Django (I’m not sure why I even chose Django in the first place). Using Tensorflow raised a lot of issues with hosting, and as such, I’d probably switch frameworks and create a new classifier.
Python, Swift, Django, Tensorflow