{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/triple-consistency-loss-for-pairing","title":"Triple consistency loss for pairing distributions in GAN-based face synthesis","arxiv_id":"1811.03492","date":"2018-11-08","proceeding":null,"authors":["Enrique Sanchez","Michel Valstar"],"abstract":"Generative Adversarial Networks have shown impressive results for the task of\nobject translation, including face-to-face translation. A key component behind\nthe success of recent approaches is the self-consistency loss, which encourages\na network to recover the original input image when the output generated for a\ndesired attribute is itself passed through the same network, but with the\ntarget attribute inverted. While the self-consistency loss yields\nphoto-realistic results, it can be shown that the input and target domains,\nsupposed to be close, differ substantially. This is empirically found by\nobserving that a network recovers the input image even if attributes other than\nthe inversion of the original goal are set as target. This stops one combining\nnetworks for different tasks, or using a network to do progressive forward\npasses. In this paper, we show empirical evidence of this effect, and propose a\nnew loss to bridge the gap between the distributions of the input and target\ndomains. This \"triple consistency loss\", aims to minimise the distance between\nthe outputs generated by the network for different routes to the target,\nindependent of any intermediate steps. To show this is effective, we\nincorporate the triple consistency loss into the training of a new\nlandmark-guided face to face synthesis, where, contrary to previous works, the\ngenerated images can simultaneously undergo a large transformation in both\nexpression and pose. To the best of our knowledge, we are the first to tackle\nthe problem of mismatching distributions in self-domain synthesis, and to\npropose \"in-the-wild\" landmark-guided synthesis. Code will be available at\nhttps://github.com/ESanchezLozano/GANnotation","url_abs":"http://arxiv.org/abs/1811.03492v1","url_pdf":"http://arxiv.org/pdf/1811.03492v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"triple-consistency-loss-for-pairing","repo_url":"https://github.com/ESanchezLozano/GANnotation","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"attribute","task_name":"Attribute"},{"task_slug":"face-generation","task_name":"Face Generation"},{"task_slug":"face-to-face-translation","task_name":"Face to Face Translation"},{"task_slug":"translation","task_name":"Translation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}