Diffusion Models Beat GANs on Image Synthesis

<!doctype html> <html lang="en"> <head>  <meta charset="utf-8"> <meta name="viewport" content="width=device-width, initial-scale=1, shrink-to-fit=no"> <title>Diffusion Models Beat GANs on Image Synthesis</title> <link rel="stylesheet" href="/static/papers/css/papers.css"> <meta name="citation_title" content="Diffusion Models Beat GANs on Image Synthesis"> <meta name="citation_author" content="Dhariwal, Prafulla"> <meta name="citation_author" content="Nichol, Alexander"> <meta name="citation_journal_title" content="Advances in Neural Information Processing Systems"> <meta name="citation_volume" content="34"> <meta name="citation_firstpage" content="8780"> <meta name="citation_lastpage" content="8794"> <meta name="citation_pdf_url" content="https://proceedings.neurips.cc/paper_files/paper/2021/file/49ad23d1ec9fa4bd8d77d02681df5cfa-Paper.pdf"> <meta name="citation_publication_date" content="2021-12-06">   <link rel="stylesheet" href="https://stackpath.bootstrapcdn.com/bootstrap/4.4.1/css/bootstrap.min.css" integrity="sha384-Vkoo8x4CGsO3+Hhxv8T/Q5PaXtkKtu6ug5TOeNV6gBiFeWPGFN9MuhOf23Q9Ifjh" crossorigin="anonymous"> <link href="/static/menus/css/menus.css" rel="stylesheet" id="bootstrap-css"> <link rel="stylesheet" href="https://use.fontawesome.com/releases/v5.8.1/css/all.css" integrity="sha384-50oBUHEmvpQ+1lW4y57PTFmhCaXp0ML5d60M1M7uH2+nqUivzIebhndOJK28anvf" crossorigin="anonymous"> <script type="text/javascript" async src="https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.7/latest.js?config=TeX-MML-AM_CHTML"> </script> <script type="text/x-mathjax-config"> MathJax.Hub.Config({ "tex2jax": { "inlineMath": [["$","$"], ["\$","\$"]], "displayMath": [["\\[","\\]"]], "processEscapes": true } } ); </script> <style> @media (prefers-color-scheme: dark) { body { background-color: #333; color: #eee; } } .btn-spacer { margin: 2px; } .footer { position: fixed; left: 0; bottom: 0; width: 100%; background-color: #eee; color: black; } </style> </head> <body> <nav class="navbar navbar-expand-md navbar-light bg-light"> <button class="navbar-toggler" type="button" data-toggle="collapse" data-target="#navbarToggler6" aria-controls="navbarToggler6" aria-expanded="false" aria-label="Toggle navigation"> <span class="navbar-toggler-icon"></span> </button> <div class="collapse navbar-collapse" id="navbarToggler6"> <a class="navbar-brand" href="/">NeurIPS Proceedings</a> <ul class="navbar-nav mr-auto mt-2 mt-md-0"> <li class="nav-item"> <a class="nav-link" href="/admin/login/?next=/admin/"><i class="fas fa-sign-in-alt" title="Login"></i></a> </li> <li class="nav-item"> <a class="nav-link" href="/admin/logout/?nextp=/admin"><i class="fas fa-sign-out-alt" title="Logout"></i></a> </li> </ul> <form class="form-inline my-2 my-lg-0" method="get" role="search" action="/papers/search"> <input class="form-control mr-sm-2" type="text" name="q" placeholder="Search" aria-label="Search" id="navsearch"> <button class="btn btn-outline-success my-2 my-sm-0" type="submit">Search</button> </form> </div> </nav> <br> <div class="container-fluid"> <div class="col"> <h4>Diffusion Models Beat GANs on Image Synthesis</h4> <p> Part of <a href="/paper_files/paper/2021">Advances in Neural Information Processing Systems 34 (NeurIPS 2021) </a> </p> <div><a href="/paper_files/paper/12295-/bibtex" class="btn btn-light btn-spacer">Bibtex</a> <a class='btn btn-light btn-spacer' href='/paper_files/paper/2021/file/49ad23d1ec9fa4bd8d77d02681df5cfa-Paper.pdf'>Paper</a> <a target="_blank" href='https://openreview.net/forum?id=AAWuCvzaVt' class='btn btn-light btn-spacer'>Reviews And Public Comment »</a> <a class='btn btn-light btn-spacer' href='/paper_files/paper/2021/file/49ad23d1ec9fa4bd8d77d02681df5cfa-Supplemental.pdf'>Supplemental</a></div><br> <h4>Authors</h4> <p><i>Prafulla Dhariwal, Alexander Nichol</i></p> <h4>Abstract</h4> <p>We show that diffusion models can achieve image sample quality superior to the current state-of-the-art generative models. We achieve this on unconditional image synthesis by finding a better architecture through a series of ablations. For conditional image synthesis, we further improve sample quality with classifier guidance: a simple, compute-efficient method for trading off diversity for fidelity using gradients from a classifier. We achieve an FID of 2.97 on ImageNet 128$\times$128, 4.59 on ImageNet 256$\times$256, and 7.72 on ImageNet 512$\times$512, and we match BigGAN-deep even with as few as 25 forward passes per sample, all while maintaining better coverage of the distribution. Finally, we find that classifier guidance combines well with upsampling diffusion models, further improving FID to 3.94 on ImageNet 256$\times$256 and 3.85 on ImageNet 512$\times$512.</p> </div> </div> <script src="https://code.jquery.com/jquery-3.6.0.min.js" integrity="sha256-/xUj+3OJU5yExlq6GSYGSHk7tPXikynS7ogEvDej/m4=" crossorigin="anonymous"></script> <script src="https://cdn.jsdelivr.net/npm/popper.js@1.16.0/dist/umd/popper.min.js" integrity="sha384-Q6E9RHvbIyZFJoft+2mJbHaEWldlvI9IOYy5n3zV9zzTtmI3UksdQRVvoxMfooAo" crossorigin="anonymous"></script> <script src="https://stackpath.bootstrapcdn.com/bootstrap/4.1.3/js/bootstrap.min.js" integrity="sha384-ChfqqxuZUCnJSK3+MXmPNIyE6ZbWh2IMqE241rYiqJxyMiZ6OW/JmZQ5stwEULTy" crossorigin="anonymous"></script> <script> $('.dropdown-menu a.dropdown-toggle').on('click', function (e) { if (!$(this).next().hasClass('show')) { $(this).parents('.dropdown-menu').first().find('.show').removeClass("show"); } var $subMenu = $(this).next(".dropdown-menu"); $subMenu.toggleClass('show'); $(this).parents('li.nav-item.dropdown.show').on('hidden.bs.dropdown', function (e) { $('.dropdown-submenu .show').removeClass("show"); }); return false; }); </script> <br> <div class="modal fade" id="myModal"> <div class="modal-dialog"> <div class="modal-content"> <div class="modal-header"> <h4 class="modal-title">Name Change Policy</h4> <button type="button pull-right" class="close" data-dismiss="modal" aria-hidden="true">×</button> </div> <div class="modal-body"> <p>Requests for name changes in the electronic proceedings will be accepted with no questions asked. However name changes may cause bibliographic tracking issues. Authors are asked to consider this carefully and discuss it with their co-authors prior to requesting a name change in the electronic proceedings.</p> <p>Use the "Report an Issue" link to request a name change.</p> </div> <div class="modal-footer"> </div> </div> </div> </div> <footer class="footer"> <div class="btn-spacer"> <a href="https://neurips.cc/Help/Contact?select=Conference">Report an Issue</a>    |    <a href="#" class="modal-link" data-toggle="modal" data-target="#myModal">Name Change Policy</a> </div> </div> </footer> <div class='hidden'>Do not remove: This comment is monitored to verify that the site is working properly</div>  </body> </html>

CINXE.COM

Diffusion Models Beat GANs on Image Synthesis