Software promises to improve AI object recognition

This is the claim of neuroscientists in the US who have designed a model that mirrors human visual learning.

Published in Frontiers in Computational Neuroscience, Maximilian Riesenhuber, PhD, professor of neuroscience, at Georgetown University Medical Center in Washington DC, and Joshua Rule, PhD, a postdoctoral scholar at UC Berkeley, California explain how the new approach improves the ability of AI software to quickly learn new visual concepts.

Machine vision system helps robots get picky

"Our model provides a biologically plausible way for artificial neural networks to learn new visual concepts from a small number of examples," Riesenhuber said in a statement. "We can get computers to learn much better from few examples by leveraging prior learning in a way that we think mirrors what the brain is doing."

Humans can quickly and accurately learn new visual concepts from sparse data, sometimes just a single example. Three-to four-month-old babies learn to recognise zebras and distinguish them from cats, horses, and giraffes but computers typically need to "see" many examples of the same object to know what it is, Riesenhuber said.