As India is encountering an alarming surge in the COVID-19 cases in the recent months, It almost became impossible for the government to keep a track of people suffering/recovered from the virus.
Hospitals are not able to provide with the required resources and are also failing in keeping a track of patients.
Plus people tend to violate the quarantine norms quite a lot. That makes it important to distinguish between a person who is COVID Positive or not and that too in the most efficient way.
Interspace tackles this exact same issue.
What it does
Interspace is a Deep Learning facial recognition app that detects whether the person is COVID-19 Positive or negative with just his/her image file.
Our Infra is capable enough to work in small Hospitals, Schools and Universities and provide them with real time data about the COVID-19 Status within their faculty.
We provide a dashboard where the administration can enter the details of their employees, students (COVID Status, Name, Address, Photos). We further use that data to train a facial recognition classifier that predicts whether the given person is COVID Positive or negative.
How we built it
We've used the dlib's state-of-the-art face_recognition libraries. Special thanks to https://github.com/ageitgey for their awesome and easy to use facial_recognition libraries!
Here, the Frontend dashboard built using React and NextJS, takes in all the data and saves it in a CSV file format.
The CSV is feeded to a MongoDB Database Server.
We used Flask to make simple API Routes to connect to the MongoDB server.
Interspace, the main facial_recognition program gets the data from the API and trains a classifier(This is only a one time process).
We then supply Interspace with a test image, it tests it with the classifier and compares it with the data from the MongoDB Server and tells if the given person is COVID Positive or not.
Challenges we ran into
The biggest challenge that we faced was while researching about the facial_recognition models.
We tried and tested many algorithms and found that the dlib one was the most accurate (95%+ accuracy) and implemented it on Interspace
The other big challenge that we faced was to manage time while building the entire full stack app within 24 hours, with the entire API and a working ML Model.
Accomplishments that we're proud of
This has been an amazing project to work on. We successfully made an entire Full-Stack App with React and NextJs as our frontend!
We used one of the most efficient face_recognition library and got more experience on working with full stack projects and Deep Learning algorithms!
What's next for Interspace-Beta
We have some big plans for the future of Interspace-Beta! We want it to integrate it with other COVID based apps to make it more efficient!
We also want to integrate automatic scraping for scraping the internet for COVID related data.