EyeDentify

Testing 2 sets of eyes.
Additional testing with 2 sets of eyes.
This is our Roboflow workflow for the model.
Blink tracker website.

Inspiration

The average screen time for teens hovers around 7 hours per day. In today's digital world, prolonged computer use has become a daily norm, leading to eye strain and related visual discomfort for millions. Our team was driven by the need to address this growing issue and enhance overall eye health in the digital age. Recognizing that most people aren't aware of the subtle signs of eye strain until it becomes a bigger problem, we developed EyeDentify. Our goal is to create a tool that actively monitors and detects eye strain, empowering users to take proactive steps to maintain eye comfort and well-being during computer use.

What it does

EyeDentify is an eye strain detection web application designed to monitor users' eye movements and detect signs of eye strain or related conditions like redness, watering, or fatigue. By attempting to combine both a trained machine learning model from Roboflow and a blink detection system, EyeDentify analyzes the user’s eye movements and visual patterns, alerting them with popup notifications when it detects strain or discomfort. The application runs seamlessly in the background, sending users real-time notifications, ensuring they take necessary breaks to maintain eye health.

How we built it

Our application is designed to process real-time data from the user’s webcam to detect signs of eye strain. The first step in our process was to gather a large dataset from Kaggle to train a custom model using Roboflow to identify patterns like redness, eye bags, and watery eyes. We were able to develop our model through implementing a workflow that divided the large scale project into smaller steps, making it easier for development. On the other hand, the front end was built with React.js, featuring a live webcam feed that will eventually integrate with our model. User authentication was implemented with Auth0 to ensure privacy to each user. While we weren’t able to fully implement the model into the website during this hackathon, that’s our next step.

Challenges we ran into

Attempting to implement our original idea, a vein detection model, proved quite challenging as we weren't able to acquire enough data to train the model. So, we decided to pivot to another solution to medical issues -- helping reduce eye strain. For this project, one of the biggest challenges we faced was gathering enough data to train our model. We initially struggled to find enough labeled data for Roboflow, which meant we had to change our approach and focus on detecting specific patterns such as redness and bagginess. Another tough part was integrating the model with our React website; we had to figure out how to make the API calls work seamlessly without messing with the rest of the website. However, we learned a lot about how to make things work despite these hurdles!

Accomplishments that we're proud of

We are really proud of creating a tool that tackles a growing problem—eye strain from too much screen time. By building a model that can track and detect eye strain in real time, we’re giving users a practical, easy-to-use solution. What makes it even more exciting is that it works entirely through the web, with no need for extra hardware, making it accessible and convenient. It helps hackers like ourselves stay informed on when to take breaks especially in hackathons like Boilermake!

What we learned

Throughout the project, we sharpened our debugging skills to make sure our model performed accurately across all the parameters we set. There were plenty of moments where things didn’t work as expected, but taking the time to troubleshoot and refine our approach helped us push through. We also found that regularly checking in on each other's progress kept us aligned with our goals and made collaboration much more efficient. By working together and staying organized, we were able to tackle challenges more effectively and keep the project moving forward.

What's next for EyeDentify

As we move forward, we plan to combine our blink tracking system with the strain detection model in the future. We’re also focused on improving the model’s accuracy by collecting more diverse data to better capture a wider range of eye strain indicators. Additionally, we plan to add features such as implementing MongoDB, so we can track more in-depth data for our users. Finally, another key goal is optimizing the app’s performance across different devices, especially for mobile users who want to keep tabs on their eye health while on the go.