Cao et al., A Comprehensive Survey on Geometric Deep Learning

Presentation can be found here.

Surveys current (as of 2020) methods used for geometric deep learning as well as a list of datasets with links, and some pointers on open problems. Also gives theoretical background but not as well as Bronstein et al., 2016. For graphs both spectral and spatial methods are covered and it is noted that spatial methods (which simply aggregates neighbor features) are winning out over more complicated spectral ones. For meshes, the focus is on patch based methods (which seem quite complex) or point based methods like PointNet. Method explanations are a bit superficial and often hard to follow, but it is a good list of what’s been tried so far. One could also see Xiao et al., 2020 for further overview.