Looking up disease symptoms online brings with it a lot of uncertainty in possible illnesses. Without the opinion of a medical professional, finding out what's wrong with you on sources like WebMD is a bit of a gamble. We wanted to bring the power of professional diagnoses into the pockets of millions. To start off, we began with a simple implementation that was also timely for the incoming Summer season.
What it does
DermaCare enables users to take pictures on their phones of strange skin, which is then analyzed using a Machine Learning model that is trained on over 10,000 images of various skin diseases. This analysis is then sent to a personal doctor that will assist in making an accurate diagnosis seamlessly. The Android app is geared towards users, whereas the website enables medical professionals to easily view patient submissions
How we built it
We used multiple Google Cloud Platform technologies, such as Cloud Storage, Cloud Functions, Firebase, and AutoML Vision to create a fully functioning and custom Machine Learning model that easily communicates with our Android app and website.
Challenges we ran into
Preparing the massive data set of images provided by Harvard University for training the Machine Learning model was a hefty task that required the creation of multiple Python scripts to organize the data in a manner that the AutoML Vision platform could understand. Additionally, the Android app was developed using Flutter, which is a relatively new SDK that lacked a lot of documentation, which presented us with many challenges. Ultimately, in the end, both web and mobile apps thankfully worked as desired.
Accomplishments that we're proud of
As an all-freshman team, we successfully learned to use all the APIs, dependencies, and platforms that we wanted to use to achieve our goal of enabling users to get fast skin disease diagnoses without having to gamble on the accuracy of online sources like WebMD that will sometimes needlessly worry users over benign symptoms.
What we learned
What's next for DermaCare
We want to expand support to a multitude of diseases while improving the Doctor's Portal. By increasing the range of known diseases we can improve the accuracy of the ML algorithm. Improving the Doctor's Portal would allow more quick and private diagnoses by trained professionals.