E-learning granted European and non-European students access to unprecedented learning opportunities. The global COVID-19 pandemic, however, turned e-learning from an opportunity to an urgent necessity. Advanced e-learning tools are therefore in high demand and are likely to remain so.
During stationary classes, for example, teachers were able to monitor their students’ focus in real-time
But how can we make sure students remain interested during online classes on e-learning platforms?
How can teachers make sure they are being listened to?
Our solution to this problem is Foxus: Student focus monitoring system
Foxus provides teachers and students with a lightweight, simple, non-intrusive tool using machine learning techniques in order to monitor user focus and discreetly provide constructive feedback, ultimately helping make lectures more effective, more interesting, and less time consuming for everyone.
Foxus monitors nearly 60 key points on the human face in order to provide real-time focus as the output.
Foxus notices, for example: concentration, fatigue, happiness, enthusiasm, curiosity and engagement.
Foxus provides more comprehensive after-class analysis.
The only thing Foxus needs is a notebook and a camera.
Our solution is fully GDPR compliant and takes student privacy seriously.
- React (allows for creation of a mobile app in the next step)
- Material-UI Design (components tested for their utility and accessibility used by major IT companies like Google)
- Communication via Websockets (real-time data streaming technology for research)
- SVG animation for explaining our analytics data
- Python Flask (microframework)
- Rest API & Websockets (technologies used for client-server communication)
- SQLAlchemy
- OpenCV (image recognition)
- Deep Neural Network with Convolutional Neural Network (user data classification)
- Neural network designed and pre-trained by the team
- Docker containerization
- managed by GKE (Kubernetes)
- EEG (electroencephalography)
- fMRI (functional magnetic resonance imaging)
-
the biggest issue for us was designing the algorithm in a way as to correctly identify collections of key points needed for focus and emotion recognition.
-
we had to change our approach to backend side of development as well as video script.
-
we had to contact a mentor in order to make sure that our project is GDPR compliant.
Our demo is deployed in the United States, therefore it can work slower than usual, unfortunately we did not have enough time to deploy it in Europe. In order to run it, you will have to turn your camera on. Please be patient.
- we succeeded in correctly identifying all the critical points we needed for our analysis.
Real-time smile data analysis
-
we managed to solve internal communication issues and successfully share complex scientific and technological knowledge across the entire team.
-
our production quality and speed increased significantly during development.
-
we have learned that things that may look simple on the surface may in fact be very multi-layered and time-consuming.
-
we have learned much about new tools (Docker, Websockets for example), techniques and solutions.
-
also: how to manage tight schedules in order to include power naps ;-)
- clinical tests in order to optimize our model's parameters with the assistance of Department of Neurobiology @ Jagiellonian University, Kraków, Poland.
- optimizing overall performance
- implementation of additional tracking functionalities like background disruption, voice analysis
- SaaS in order for Foxus to work with all of the most popular video communicators on the market:
- introduction of the software as not only education, but also business-oriented solution (team meetings, internal trainings)