-
-
When students are successfully social distancing, they will have a green box around them and no warning will be notified to the teacher
-
When a lot of students are in red there will be a warning sign around the cluster of students not distancing and teacher will see alert
-
This is the UI of the mask detection system that we are currently developing and training.
-
When students are not wearing their masks, the app will detect if students are not wearing their mask
-
If students are not complying with the guidelines, an alert will be sent to the teacher notifying them that students are not distancing
Inspiration
As schools reopen amidst the COVID-19 pandemic, students must social distance to prevent themselves and teachers from getting ill . According to The New York Times, new studies reveal that “ children younger than age 5 may host up to 100 times as much of the virus in the upper respiratory tract as adults.” This goes against the widespread belief that children are safe from the virus, but in reality, they are susceptible to transmit and acquire the virus. Schools need to take more precaution. However, it will be difficult to monitor each student, especially when the teacher is busy teaching. We know so many people who work in the education field, including close relatives, so this is a serious concern to our team. We want to help our community stay safe by preventing spread, so we decided to make an application that would help teachers avoid COVID-19 spread in classrooms . Source: https://www.nytimes.com/2020/07/30/health/coronavirus-children.html
What it does
With technology, we can solve this problem by detecting if students are less than 6ft apart and alert the teacher, and show the region in the classroom where students are not complying . We also included a way for the app to detect if students are wearing their masks correctly and the teacher will be notified if students are not . With these analytics, the teacher can successfully alter the classroom environment if needed and enforce the rules. This will significantly prevent students from getting and spreading COVID-19 and prevent teachers (many of whom are more susceptible) from getting COVID-19 and protect our community .
How we built it
We used python to build the framework of the app. We explored OpenCV , YOLO , and Darknet to build the image recognition feature of detecting people and detecting if they are 6ft apart. We also used Linux, we used YOLO software, OpenCV, and CUDA to start training and testing our machine for face mask detection. While we did get some preliminary code working, we do not have a functioning product for face mask detection. Our face mask prototype in the demo simply showcases the UI we want to achieve. We are still working on training and testing our machine.
Challenges we ran into
It was our first time using an open source neural network so figuring out how to connect these components in such little time was difficult. We had to go through a lot of documentation to learn about this technology and then it took hours to implement in our application. It was also difficult to be able to figure out the average area among all the “red zones” using OpenCV, but we are pleasantly surprised that we got it to work with great accuracy. In terms of mask detection, we had a lot of difficulty using Linux to utilize YOLO software, OpenCV, and Darknet. While we did get preliminary code to work, we are still working on training and testing our machine.
Accomplishments that we're proud of
We are so surprised that we were able to successfully implement image detection from a neural network this quickly. Although the application isn’t as seamless as we would have liked, the components that we wanted worked and this is exciting. We got the distance tracker to predict people who are under 6ft apart working, person detection working, and got it to analyze if people are wearing their masks successfully.
What we learned
We learned how to use neural networks to process image detection using YOLO and OpenCV. We are both in the process of learning python as well so taking this leap to learn this advanced technology was stressful, but we got a lot out of the process. We also explored the real-time image detection capabilities of OpenCV and had to learn the syntax to use its components in our application
What's next for CovidOut
We are planning to combine these components that we built into one seamless application for schools to use . We hope to publish this software in the app store soon before the school year begins. We want to make the UI more user-friendly and interactive (navigation, clear analytics page) as possible so we will continue to develop that. We can envision this app to significantly help save lives this school year and keep communities nationwide safe .
Log in or sign up for Devpost to join the conversation.