Dermi-Dilemmas

Inspiration

When James had a rash on his upper thigh he didn't know what to do and took a photo of the rash to send to his mum. His mum, a nurse, helped him through the process - and always has. The inspiration for this project was to help people not so fortunate as James who might not have a doctor in the family.

What it does

The user will upload an image of their skin disease, and if the machine learning model determines they have one of the 9 skin diseases we are checking for, then it will highlight their problem, give them an overview of what they have, and provide an NHS link for more information.

How we built it

James and Dongyiu worked on the website and API connections, whilst Ellie and Rishi worked on the machine learning model using Tensorflow.

Challenges we ran into

The dataset we used has bias (images only include white people) and has limited data (only 9 categories of skin disease). If we had a larger database, with high quality images then our model would be much more robust, and produce a much more accurate results.

Accomplishments that we're proud of

User friendly website
Reasonably precise diagnoses (75%)
Accurate, researched information

What we learned

Teamwork, communication, planning and problem solving. All of us in the team learnt more technical skills in their respective areas.

What's next for Dermi-Dilemmas

Creating a chat bot to go along with the model so it can prompt users with questions just like doctors.
Separating each skin disease into severity so that our model can give a more personalised response.