Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks · HackerPapers discussions

Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks — mentioned by jeeceebees

Fri, 10 May 2019 11:29:37 GMT

Would be even nicer if it could be trained on unpaired datasets (ala CycleGAN https://arxiv.org/abs/1703.10593).

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

Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks — mentioned by soraki_soladead

Tue, 01 Jan 2019 00:54:35 GMT

"Overfitting" would be a little more specific. Basically the task was easier to accomplish by memorizing the data than by generalizing. This was pretty clear from the original CycleGAN paper[0] (figure 4; bottom row) since it was generating trees in the exact positions of the source image where the map had no indication of tree placement. 0. https://arxiv.org/pdf/1703.10593v1.pdf

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

Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks — HN submission

Thu, 04 Jan 2018 15:38:22 GMT

Unpaired Image-To-Image Translation Using Cycle-Consistent Adversarial Networks

1 point; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.

Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks — mentioned by aisofteng

Sat, 01 Apr 2017 06:01:29 GMT

Style transfer lies outside my specialization, so take my comment as speculation informed by intuition from variously related works that may not correctly carry over to this one. If you look at the nature of the transformations achieved in this paper, you might note that they are changes in, well, style. That is, the presentations of the objects in an image are represented using a different _style_, but they remain the same object. As an example, take a look at the horse/zebra transformation; the horse obtains a zebra's stripes, but it _structurally_ still looks like a horse. That is to say…

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

Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks — mentioned by wonderous

Fri, 31 Mar 2017 15:17:24 GMT

Direct link to research paper: https://arxiv.org/pdf/1703.10593.pdf

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