As we know that digital education is being promoted across the world. it is growing at a CAGR of 8.25% from 2020 to 2027. With new technology comes new problems, i.e. for a teacher, parent and, student and we have tried to solve these problems.
As we all know due to this pandemic (COVID-19) we all know how much it is important for us to shift toward online learning platforms or digital education.
So, In online classes/meetings, it becomes difficult for a teacher to know if all students are attentive or not.
Similarly, we know that reading course books become quite difficult and time taking,for many people it becomes very hard to go through the textbook. Also, students found it difficult to understand languages other than their native language, so audio translation is a very important aspect of learning.
Therefore we have to build a model to solve the above mentioned problems.
What we learn?
With this project, we learned a lot about online team management, working constantly towards solutions, and debugging whenever required. It was difficult for us to work under pressure but everything remains under control. Also, we learned about some new technologies like bubble.io and echoAR although not able to use any of them.
We learned a lot about OpenCV and face detection in this project. Also learned about text translation and conversion of different file formats in audio.
How we built?
- We have made a Django application for testing purposes that can be used as a plugin or extension in the future. We have used python libraries and opensource model data to achieve our results. Also, we have included our open-sourced python package(freshlybuiltimagebol) that we are working on in this project.
Problems we faced
It becomes difficult at first to get a working model for attention span detection. We searched for training data to make a model and luckily found some. We tried to include some more things like electronic device recognition( like phone detection etc). Although the model is working, its accuracy at the current movement is not that good.
Also, we tried to include a natural handwritten text recognition but it results in futile.
- We can improve model accuracy and build a browser extension for span detection and audio conversion.
- A handwritten text recognition model can be added to improve audio conversion results.
Log in or sign up for Devpost to join the conversation.