The Post-it® app uses a custom computer vision engine to detect Post-it® Notes placed on a surface. It can then extract and enhance the notes allowing the user to continue working digitally.
The engine is tested and improved using a dataset of images, each representing a variable number of notes from different angles. This dataset contains high quality real-world samples, but it is limited in size, and requires a tedious hand-labelling procedure for each image.
An interesting solution to this problem could be to generate such test images. Not only would this eliminate the need for hand-labelling, but it would also open for new test scenarios and enable the use of new metrics to track the performance of the capture engine.
In this project, you will use a 3D rendering engine of your choice (e.g. Unity), and experiment with different techniques to generate photorealistic images of notes placed on a surface. This component could then be connected with the Post-it® scanning engine to evaluate the quality of the renderings, and investigate how different parameters (such as noise, distortions, lighting, etc.) affect the note detection.