The lack of real-time analysis of two-way virtual communication, a technology which has nowadays become the fore-runner in the field of smart education, hampers the efficient conversation which should happen between a teacher and students. So, there has always been a dire need for a good solution for such a problem, not only to make our current system of education better but also to make it more feedback-oriented, thus, having a larger degree for healthier development of the children. So, we have built a Project which revolves around the "Smart Education" theme of Hack36. It aims at improving the quality of E-education over a local network, through one-to-many Video communication, i.e., teacher-to-students live classes in a local college or university network. We are making it more analysis-oriented by using in it technologies like Facial Detection using Deep Learning, real-time Feedback using Charts and Alert systems, etc. We are also creating a summary at the end of the video lecture, both for the teacher's and the students' sides. At the end of the session, based on the data received by the ML model, we can show the general attentiveness of all the students in the class to the teacher, highlighting the duration of the lecture where the students felt the most boredom or where most students had doubts, giving the lecturer an analysis on his/her performance and helping him in further improving the course. On the Students' side, he/she would receive the feedback about his/her own concentration for the whole duration. Besides these features, the students and teachers, both would have a dashboard - where their past performance would be stored and also, they can add a to-do list for them to finish for a particular day.
We needed raw frames from the video conferencing platform to process it. However, there was no feasible service that provided us with this. Thus we build our own video conferencing platform using WebRTC which supported one too many broadcasting. For analyzing frames in realtime, we had to build a separate API server and had to optimise dlib model for our needs. We had to store the data of each student and each room and organize it such that it was easy to access and it can be provided to the teacher in real-time.
Discussion