The Post-it® app allows users to capture physical Post-it® notes and continue to work with them digitally on a smartphone or tablet.
A typical scenario is that of Kanban boards, where the user writes different tasks on separate notes, and then places the notes on a board divided into columns (e.g. “to do”, “in progress”, “done”). For this particular use case, the position of the notes matters, and it is bound to change over time as tasks move along the columns.
The current version of the Post-it® app does not keep track of which notes have already been captured, and does not extract any logical meaning from the position of the notes. In the example of the Kanban board, this means that users that capture different snapshots of the same board will end up with duplicate notes.
In this project, you will investigate different methods to identify Post-it® notes that appear in different pictures, in such a way that duplicates can be avoided. The implementation should also be capable of tracking the identified notes between different captures, trying to extract meaning from the changes in position. You will be free to choose the methods you would like to experiment with, including old-school Computer Vision algorithms, Machine Learning, or a custom solution.