Plenty of robots are capable of recognizing objects and people. Researchers are always looking for ways to speed it up though. While deep learning techniques can help, they have limits in robotic applications. Elisa Maiettini, Giulia Pasquale, Lorenzo Rosasco, and Lorenzo Natale have come up with a new approach to speed up object recognition in robots.
The researchers used the FALKON algorithm, which is a kernel-based method for fast training on large scale problems. You can find out more about this approach here.