Imitation learning supported by language for industrial robotics
Project Dataset Team Links
We consider the problem of visual imitation learning from human demonstration, where we wish to learn a 6 degree of freedom (DOF) trajectory of an object from RGBD video capturing a human performing a complex object manipulation task. Reliably estimating and tracking the 6DOF position of the object remains a key open challenge as objects often lack distinctive texture, have symmetries, and are occluded by human body.
Project is funded by Technological agency CR, project No. TJ01000470. Principal investigator Karla Stepanova, CIIRC CTU. From October 2017 to September 2019.