HTC Seminar Series #31: Geometric Deep Learning

 Today's HTC seminar series is a talk titled 'Geometric Deep Learning, from Euclid to drug design', presented by Michael Bronstein virtually at Imperial College London.  This is an awesome lecture.  Grab a big cup of coffee, sit back, and take it all in.


Observations:

1: Geodesic CNN's, super cool.  Geodesic image processing is a little known thing, but very useful.  All over Studio Artist for example.

2: Note the discussion of priors in these systems.  How the built in priors give you various kinds of symmetry transform invariance. 


Comments

Popular posts from this blog

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

Pix2Pix: a GAN architecture for image to image transformation

Smart Fabrics