We came across some startling statistics about the dangers of skin cancer, including melanoma. Approximately 2.3 percent of men and women will be diagnosed with melanoma of the skin at some point during their lifetime. So many people suffer from melanoma, but we noticed much less attention and help efforts went towards melanoma compared to other diseases.
What it does
SkinSafe provides deep learning analysis on moles, straight from people's phones. Through over 10,000 training images, our AI can detect the presence of melanoma without the need of a doctor. Just snap a photo of your skin and that’s it!
How we built it
The frontend of the app is built with React Native, and it runs on iPhones and M1 Macbooks. The app includes features like a built-in camera, interactive map, and offline access to our machine learning model. Our model was composed of deep learning architecture; specifically a Convolutional Neural Network, an Image based deep learning model. It was trained with an open source dataset consisting of over 10,000 images.
Challenges we ran into
The biggest technical challenge the team ran into was connecting the machine learning model with the users' phones. We tried different methods, including hosting the model on a server and uploading images to the cloud, but we ultimately decided to run the model on the phone itself.
Accomplishments that we're proud of
One thing that we're proud of is building an actual working product. This app can be installed on real phones and it works. Our project could legitimately help victims of melanoma.
What we learned
We learned the process of building an app with multiple people working on it. Every time we changed something, we went through the whole app again and re-tested everything to make sure nothing broke.
What's next for SkinSafe
There are many possibilities for growth in the future, too. We could hone our AI to detect which type of melanoma a person has, which can further help patients. We could also partner with doctors, so doctors could tell their patients to try the app first before booking an appointment, which would increase our product’s usage.