An Intriguing Failing of Convolutional Neural Networks and the CoordConv Solution · HackerPapers discussions

An Intriguing Failing of Convolutional Neural Networks and the CoordConv Solution — mentioned by nmca

Sat, 11 Aug 2018 11:40:39 GMT

Hm, this seems vaguely interesting-if-true but: 1) It's not published in a good CV journal. 2) It seems quite likely that they've just identified the problems described and fixed by [0, 1] (of which [0] is the much better paper). In brief, CNNs guarantee translation invariance (ignoring boundary effects) through translation equivariance of their feature maps; this borks certain problems, most importantly a commonly used mechanism for instance segmentation as described in [0]. Consider trying to instance-segment two copy/pasted faces; translation equivariance and global pooling clearly make it…

Paper and discussion history. A mention is not an endorsement.

An Intriguing Failing of Convolutional Neural Networks and the CoordConv Solution — mentioned by ebalit

Sun, 22 Jul 2018 12:32:52 GMT

From the abstract: "For any problem involving pixels or spatial representations, common intuition holds that convolutional neural networks may be appropriate. In this paper we show a striking counterexample to this intuition via the seemingly trivial coordinate transform problem, which simply requires learning a mapping between coordinates in (x,y) Cartesian space and one-hot pixel space. Although convolutional networks would seem appropriate for this task, we show that they fail spectacularly. We demonstrate and carefully analyze the failure first on a toy problem, at which point a simple…

Paper and discussion history. A mention is not an endorsement.