Problem
I. Object recognition is a challenge for people with low vision or blindness. Over the last 10 years or so, a revolution in computing has allowed the building of apps that can help. These programs use a new technology called "deep neural networks", which are very loosely based on some of what we know about visual processing in real brains. The apps are becoming useful for people in real everyday situations, as they can use their phones to take a picture of the world in front of them, and the apps can tell them, or help them figure out, what is in it. Helping improve this technology is also a great job opportunity for perceptual psychologists. Describing (i) a couple situations where you think the technology would be most useful, and (ii) a couple where you think it would fail. Justify your answers in a few sentences.
II. One theory for processing of upside-down faces is that we see the parts of the face, but don't get a good sense of the face as a whole. For the "Former President Obama" illusion , put why just processing the face parts, but not the whole, might cause us to not notice how weird the face at right is when we see it upside-down?
III. Explain why the distorted face looks OK upside-down using the Bayesian inference ideas explained in the video lectures. Specifically how can the use of priors and/or likelihood explain the misperception of the upside-down distorted face?