Do AI algorithms see the way we do? Towards a dialogue on vision between brains and machines.
Deep neural networks have revolutionised computer vision with their impressive performance on a range of tasks. Recently, their object representations have been found to closely match those in the brain's visual areas.
Yet their performance still falls short of human performance on these tasks, and it has been challenging to understand why deep networks work or how they can be improved.
Arun's lab has been comparing object representations in brains and machine algorithms to understand how we see and to make machines see better. Their lab has shown that comparing deep networks with brain representations can reveal systematic biases, and that fixing these biases improves performance.
They have also tested deep networks for a range of classic perceptual phenomena. Taken together, these results suggest that accumulated wisdom from visual neuroscience can help us understand and improve deep neural networks.
About the speaker:
SP Arun received his B.Tech from IIT Bombay, and his MS & PhD from Johns Hopkins University, all in Electrical Engineering. He completed his postdoctoral research at Carnegie Mellon University and joined as faculty at the Centre for Neuroscience.
All along the way, he read too much science fiction for his own good, got fascinated by why robots still cannot solve the most basic tasks that our brains solve so effortlessly, and began studying sensation and perception in the brain.
His lab, the Vision Lab @IISc, studies how the brain solves vision by investigating behaviour and brain imaging in humans, monkeys and by comparing vision in brains and machine algorithms.
To attend online:
Add to calendar: https://shorturl.at/cYf4Y
Gmeet link: meet.google.com/efm-ittp-wbn
Looking forward to seeing you!