Having attended Microsoft's workshop on Azure, I was immediately drawn to the custom vision API and was excited to test it out. I thought of a simple problem that I actually came across a couple of times in Reddit discussions, asking what was a common medical symptom people often overlooked -- moles! Strange growths -- that turn out to be malignant and cancerous! Often, people dismiss these things that seem minor until it's too late.
What it does
This website allows the user to upload an image of a growth they are concerned about on their skin, and will receive a recommendation on how strongly they should see a doctor, based on the probability of how closely the growth resembles melanoma.
How I built it
Using Microsoft's Custom Vision API; I uploaded several images of "regular moles" and definite instances of melanoma to train it to detect the probability of a growth or mark on skin being showing characteristics of melanoma. High probability is meant to represent the information that the growth is abnormal enough to warrant seeing a doctor to get it checked out.
Challenges I ran into
Accomplishments that I'm proud of
Having a straightforward idea that was within reach of my beginner status and being able to execute that idea to produce a product I can demo. Building this website was very much a first for me and being able to see it do the thing I want it to do feels fantastic.
What I learned
What's next for Melanoma Project
Finding databases of even more pictures in order to train my custom vision to become more accurate and determining the probability melanoma -- and hence, of encouraging a person to go see a doctor for a full body examination! On a larger scale, it would be nice to build this into something that could serve as a reminder or encourager for people to take care of their health. For instance, putting off seeing a dentist can incur very high charges if a person eventually needs a root canal after delaying a cavity check-up for too long.