This project takes as input an existing video of a recorded lecture where a black or white board is used. From the video, we plan on writing software which will detect at what points the lecturer moves away from the board (or if the lecturer rubs off the content of the board if we have time), take the images of the frames at which this happens, postprocess this to look better and put all these images in a pdf. This will mean you would be able to auto-generate "notes" from a recorded lecture without any effort.
Also, if we have time, we could add an option to insert the transcribed words of the lecture alongside the images of the board, using a 3rd party speech-to-text tool.
- Write a function which will determine the position of a lecturer in video and/or the positions where a board is rubbed off.
- Write a function which will determine the position of a board in a recorded lecture.
- Write/use an existing board post-processing function.
- Write a web front-end to use this.
- (extension) Write the speech-to-text functionality, as proposed in the description.
We already have a team of 6 people, but if you're interested in the project you can join!
Things you might need to be interested in:
- Computer Vision (for recognizing when to take a photo) in Python.
- Image processing (for processing the image of the board) in Python.
- Front end web development (for the user interface) in HTML/CSS/JS.
-