© Pint of Science, 2020. All rights reserved.
The way we see guides our behaviour. Distinguishing images and evaluating scenes are essential sensory abilities of our brain that allow us to make correct decisions in tough situations. Fight or flight? - It all depends on what you see. Come and join us to learn how important neural networks are in our visual perception and how computers use such neural networks to see what humans see.
Cracking the visual code
Tom Shallcross (PhD Student)
A fundamental aim of neuroscience is to try and understand how the visual world is represented in the brain. For example, what neural activity allows the brain to distinguish prey from predator? And how is this information used to generate the correct behaviour – hunt or escape? Come along to find out how and why zebrafish may be the answer to these questions, and how scientists are generating new insights into the basic rules which govern sensory perception and behaviour.
In order to see with your eyes, first learn to imagine with your mind
Ali Eslami (Research Scientist at Google DeepMind)
Scene understanding (a.k.a. visual perception) is a requirement for intelligent behaviour. We humans are so good at it that we don't realise just how many complex cognitive faculties are involved in the process. We also don't know how infants learn to do this by themselves. I will talk about our recent research on computers that can learn to see without any human supervision. I will show how we train a large neural network to look at an image and then predict the appearance of the scene contained in that image from any new angle, in the process learning to both 'see' and 'imagine'.