Skinteract

Inspiration

According to skincancer.org, 1 in 5 Americans will develop skin cancer by the age of 70. It's often that these problems are left untreated, ignored, or even misdiagnosed by human error. Skinteract gives you a quick and accurate method to test for these problems before they surface and thus allowing you to tackle the problem earlier.

What it does

- The user uploads or takes a photo of any given mole on their body.

- It'll then feed the image to a machine learning image classification model built in TensorFlow with a roughly 80% accuracy rate.

- The app will then output the probabilities of various types of skin cancer that may exist or develop based on the image provided.

How we built it

On the front end, node.js and React Expo are the tools we used for the UI and basic functionality. On the backend, we trained an image classifier model in Tensorflow and used it to make predictions on the input photo.

Challenges we ran into

• Downloading the dataset for the images and preprocessing them took many hours but it helps to have more data, as it results in improved accuracy.
• Connecting the backend to the frontend, since we had no previous experience in React or node.js
• The same goes for the Tensorflow model and the app itself, since we had to find the right file type for the model and find out how we would integrate it with node.

Accomplishments that we're proud of

We're proud of the fact that we used tools that were completely new to us in order to perform a task we've never done before. And also, the fact that we were actually able to get some sleep as well.

What we learned

Aside from a lot of knowledge about these development tools, we learned how we can effectively manage our team, and our time as we try to rapidly learn new concepts.

What's next for Skinteract

• Improved accuracy
• More health concerns will be scannable.
• New UI

Built With

• node.js
• python
• react
• react-native
• tensorflow