008 - Skinalytics

Inspiration

Research has shown that increasingly, people do not see the doctor due to downplaying the severity of it, or being too embarrassed to. For example, 65% of UK men cite the above reasons for not wanting to visit the doctor. Skinalytics aims to address this.

What it does

Skinalytics is a Snap N Diagnose solution that can identify over 30 common skin diseases (including STDs) through the use of deep learning. Users can access our service from both a web app and facebook messenger.

How it was built

To achieve this feat was a 3 step process.
1) We first scraped 15000 images, feature engineered by hand, and preprocessed to obtain a total of over 60000 images.
2) We researched and developed a customised Covolutional Neural Network for this task alone, and repeatedly tuned the network to ensure maximum accuracy
3) Finally, we developed both a webapp and a facebook messenger bot for user accessibility.

Challenges we met

Getting quality images to better improve accuracy

Fine tuning the model to ensure high testing data accuracy

What's next for Skinalytics

Improving functionality by incorporating additional features like severity analysis, automated appointment booking and recognizing how the disease affects different areas of the body