HTC Seminar Series #33: Probing Sensory Representations

 Today's HTC Seminar Series presentation is by Eero Simoncelli titled 'Probing Sensory Representations, and was presented at the MIT Brains, Minds, and Machines summer course in 2015.  

This is an awesome lecture, and he very clearly lays out how we can take this concept of visual metamerism, and extrapolate it from color metamerism to texture metamerism to more elaborate higher order perceptual metamerism.  And how that maps to increasingly higher order models of the human visual system.


Observations

1: I think you can push the whole visual metamerism angle much higher up the perceptual foood chain than he does in this talk.  Ask yourself why to GAN systems even work at all?  Why does the fake output look real to people?  I belive you will find the answer right here.

2: The texture synthesis work he is describing pre-dates neural style transfer by quite a bit.  I was quite familiar with it at the time it was done, but kind of missed the iterative gradient descent part for the reconstructions. A concept that should be clear to anyone looking at more recent feature visualization work in CNNs.


But of course we aren't finished here.  Today's seminar is a double header.  This next talk is on the 'Perceptual Implications of Hierachical Models'.  This talk was given at CCBM 2018.


Observations

1: Local gain control pervades biological systems, and is probably very important in a theoretical computation sense.  And you might want to think about how to add that to the neural net toolkit for deep learning systems.

I also wonder if it relates to this whole notion of skip connections being so great somehow.

2: Fascinating how both CNN models do worse than the very simple biologically inspired models for the 2 visual tasks he looks at.  The visual distortion one is particually interesting.

Comments

Post a Comment

Popular posts from this blog

CycleGAN: a GAN architecture for learning unpaired image to image transformations

Image
 Like the Doublemint Twins touting the joys of Doublemint gum, 2 GANs are surely better than 1 GAN.  Especially if we package them together inside of one meta GAN module.  And this is exactly what the CycleGAN architecture does. Have you ever harbored dark secrets of turning a horse into a zebra?  The CycleGAN was developed to do just that. Learn how to turn a horse into a zebra.  And more. Now right away you can notice a difference between the image to image transformation GAN architectures we've been discussing over the last few posts.  Those last few posts described systems that learn from a database of matched input-output image pairs. And if your goal is to turn an edge representation into a nicely filled in continuous tone image, it's easy to build your database of matched input-output image pairs that your GAN system can then learn off of.  Take a continuous tone image (which will be the output of the database pair entry), then run it through an edge detector algorithm. 
Read more

Pix2Pix: a GAN architecture for image to image transformation

Image
 I thought following up yesterday's TraVelGAN post with a Pix2Pix GAN post would be useful to compare what is going on in the 2 architectures.  Two different approaches to the same problem. I stole this Pix2Pix Overview slide below from an excellent deeplearning.ai GAN course (note that they borrowed it from the original paper) because it gives you a good feel for what is going on inside of the Pix2Pix architecture.    Note how the Generator part is very much like an auto-encoder architecture, but rebuilt using the U-Net architecture features (based on skip-connections) that fastai has been discussing in their courses for several years before it became more widely known to the deep learning community at large  (and which originally came from an obscure medical image segmentation paper) . So the Generator in this Pix2Pix GAN is really pretty sophisticated, consisting of a whole image to image auto-encoder network with U-Net skip connections to generate better image quality at highe
Read more

Smart Fabrics

Image
"What makes smart fabrics revolutionary is that they have the ability to do many things that traditional fabrics cannot, including communicate, transform, conduct energy and even grow."  This is a quote from Rebeccah Pailes-Friedman, who is a professor at Pratt with a research focus on developing wearable technology and smart textiles. Enhancing textiles can be viewed through different lens.  Enhancements can be aesthetic, or performance boosting, or durability boosting. 
Read more