Speaker: Dr. Dylan Campbell, ANU.
Tutorial website: http://eccv2020.deepdeclarativenetworks.com
TL;DR: We survey some applications of deep declarative networks, focussing on pooling, projection, and geometric model fitting layers.
Abstract
In this tutorial session, we examine several applications of deep declarative networks. We focus on the subset of applications that involve pooling (including robust pooling, and rank pooling), projection (including Lp-sphere projection, limited multi-label projection, and transport polytope projection), and geometric model fitting (including optical flow and egomotion estimation, and camera pose estimation). This diverse set of examples is by no means comprehensive (missing, for example, a significant body of work in meta-learning and control), but give a good sense of the breadth of possibilities available.
