The objective of the project is to improve accuracy of objects' descriptions generated by the RoboSherlock package in difficult environments. Glossy transparent objects can not be properly detected by off-the-shelf RGB-D cameras which use infrared structured light, like Kinect. Or as another example occluded object may be not properly classified or even segmented as one peace. So the goal of this project is to study and implement algorithms which will address such problems.


Mykola Dolhyi


  • Jan-Hendrik
  • Ferenc Balint-Benczedi