We were inspired to create this hack because many of us and our friends struggle with acne everyday. We believe that our website's quick and easy diagnosis and treatment of acne will help a lot of people.
What it does
Our hack is a website that asks a user to upload a selfie image of their acne. It then runs through an image classifier and reports the severity of acne that the user has (mild, moderate, or severe). The website then recommends certain diagnoses and treatments for the user's acne. Finally the website asks for the user's zip code. Using that information, it is able to display a map of nearby dermatologists that can treat the user's acne, with in depth information on their services.
How we built it
We knew we wanted to create this concept from the start, but choosing a platform to get started with was difficult. We were stuck between Clarifai and Google Vision, but after playing around with both, we decided to use Clarifai because of its ease and speed to train which makes it great for hackathons. We first built a custom image classifier using Clarifai, which created a neural network system that classified images of a person's face to a type of acne (mild, moderate, or severe). We trained this image classifier using an acne image database from kaggle.com. We then built a frontend in HTML that would be used to use our hack. Along the way, we learned how to implement Bootstrap 5 and CSS to make our page look appealing (React was too scary to learn in under a day). After this, we also used the Google Maps API to display a map of nearby dermatologists onto the website. Finally, we built a back end server using Express.js to call to our machine learning model and implement it into our front end to analyze the acne of users. To wrap it all up, we took advantage of our credits generously granted to us and used a Google Cloud Virtual Machine as a server to run our website on.
Challenges we ran into
Accomplishments that we're proud of
Some accomplishments that we're proud of are learning how to properly use APIs in our project, successfully training a machine learning model, building a functional and an attractive frontend, and somehow managing to learn and apply so much in such a short amount of time.
What we learned
What's next for NoMoAcne
We have big plans in the future for NoMoAcne. One of our plans is to classify a person’s acne based on the types of pimples on their face (such as whiteheads, blackheads, papules, etc). Unfortunately, due to a lack of data online, we were unable to gather enough data to train our image classifier to do this in the given hackathon timeframe. Nonetheless, we plan to look more into this idea in the future and expand the functionality of NoMoAcne to include more similar and informative features.
Go visit our website link to help us eradicate acne from the online internet!