I was honored to accept the invitation to participate in the Google Summer of Code 2014 programme. I mentored a project for OpenCV about surface-based object recognition & pose estimation done by Tolga Birdal.
This module introduces a novel object recognition and pose estimation technique. It has potential use-cases in manipulation, tabletop grasping, bin picking. My personal interest is to make it part of the Object Recognition Kitchen, and use it on the REEM robot.
The GSoC is over, and the project was successful. The module has a working example with documentation. Of course nothing is perfect, especially with such projects where there are very strict deadlines, and sometimes this can have it’s effect on the quality of the code. I’m currently in the process of cleaning it up a bit and designing a better API. Luckily there is still plenty of time until the release of OpenCV 3.0 (2014 December)!
- Drost et al first paper on Point Pair Features: http://campar.cs.tum.edu/pub/drost2010CVPR/drost2010CVPR.pdf
- Object Recognition Kitchen: http://wg-perception.github.io/object_recognition_core/