[visionlist] visionlist and CVNet - Local global motion GIF

[Saul, Alan]

Thank you so much for everybody's input. We still don't seem to have a standard name for the creator, though. I guess I need to look at Twitter, and matlab. I did look at tiny portions of Michael's great site, and agree that this one is related to many others. Shapiro's Perpetual Diamond is particularly wonderful. Blivet and Escher and all of the bistable images seem to me to illustrate the same general idea.

For my lack of money, the concept is that the local spatiotemporal quadrature motion cues lead to the (false) global percept that the Necker cubes are rotating. This dependence on local information for driving global percepts is easy to accept, yet perhaps surprising that it overcomes the expectation that there are global cues. It therefore suggests that global cues don't exist, or at least require cognitive inputs, which we can use to correct our misperception. I would very much welcome corrections to this view, but reading Braddock, Cropper, and others hasn't persuaded me otherwise. Motion capture might be one argument, but maybe Murakami and Shimojo have cast doubt on that.

If you model the brain as a sheaf, it helps to understand how we might base our learned global sections on configurations of local values of the stalks. I feel that globalization happens all the time in the brain, with the creation of direction selectivity from quadrature inputs a great example (Carney and Shadlen). Wherever local information is combined to generate novel information (emergent properties), that is globalization, even when on a small scale, so that they are reinstated as new local cues. I don't think the usual examples of globalization across _space_ are different, just a convenient base space for vision. "All cues are local."


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://visionscience.com/pipermail/visionlist_visionscience.com/attachments/20200229/b66c7c10/attachment.html>