A team of researchers from Twitter have published a paper detailing a machine learning technique that uses a generative adversarial network to make shrewd guesses about how to up-res small images by up to 400%, into crisp, large images, with eye-popping results.
Despite the breakthroughs in accuracy and speed of single image super-resolution using faster and deeper convolutional neural networks, one central problem remains largely unsolved: how do we recover the finer texture details when we super-resolve at large upscaling factors? The behavior of optimization-based super-resolution methods is principally driven by the choice of the objective function. Recent work has largely focused on minimizing the mean squared reconstruction error. The resulting estimates have high peak signal-to-noise ratios, but they are often lacking high-frequency details and are perceptually unsatisfying in the sense that they fail to match the fidelity expected at the higher resolution. In this paper, we present SRGAN, a generative adversarial network (GAN) for image super-resolution (SR). To our knowledge, it is the first framework capable of inferring photo-realistic natural images for 4x upscaling factors. To achieve this, we propose a perceptual loss function which consists of an adversarial loss and a content loss. The adversarial loss pushes our solution to the natural image manifold using a discriminator network that is trained to differentiate between the super-resolved images and original photo-realistic images. In addition, we use a content loss motivated by perceptual similarity instead of similarity in pixel space. Our deep residual network is able to recover photo-realistic textures from heavily downsampled images on public benchmarks. An extensive mean-opinion-score (MOS) test shows hugely significant gains in perceptual quality using SRGAN. The MOS scores obtained with SRGAN are closer to those of the original high-resolution images than to those obtained with any state-of-the-art method.
Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network [Christian Ledig, Lucas Theis, Ferenc Huszar, Jose Caballero, Andrew Cunningham, Alejandro Acosta, Andrew Aitken, Alykhan Tejani, Johannes Totz, Zehan Wang, Wenzhe Shi/Arxiv]
Photo-realistic single image super-resolution using a generative adversarial network [The Morning Paper]
(via Four Short Links)
Lawrence Lessig was once best-known as the special master in the Microsoft Antitrust Case, then he was best known as the co-founder of Creative Commons, then as a fire-breathing corruption fighter: in America, Compromised, a long essay (or short nonfiction book), Lessig proposes as lucid and devastating a theory of corruption as you'll ever find, a theory whose explanatory power makes today's terrifying news cycle make sense -- and a theory that demands action.
In PrinTracker: Fingerprinting 3D Printers using Commodity Scanners (Scihub mirror), a paper to be presented at the ACM SIGSAC Conference on Computer and Communications Security conference in Toronto this month, a group of U Buffalo and Northeastern researchers present a model for uniquely identifying which 3D printer produced a given manufactured object, which may allow […]
More than 40% of US federal judges have attended Manne seminars, a notionally "bipartisan" educational conference presented by a Florida "Law and Economics" institute whose invited ideological allies explained to judges why pollution is good for minorities (polluted neighborhoods are cheaper and therefore affordable by poor people), unions are bad, monopolies are economically efficient, discrimination […]
Speed reading isn’t just an innate skill possessed by a lucky few. Anyone can learn to speed read, and the benefits are endless. The brain can process more information than most people have time to soak up, but you can make that time now with the 2018 Award-Winning Speed Reading Bundle. The first half of […]
Sure, you could use the same old PowerPoint templates for your next business presentation. It’s not like you have bosses or investors to impress. Oh wait, you do? Time to augment that slideshow with Slideshop – the presentation tool that can individualize your pitch while saving you time. Compatible with PowerPoint, Keynote and Google Slides, […]
Multinational companies have used the no-nonsense methodologies of Six Sigma and Lean Six Sigma to oil a smooth-running operation for years. What is it? Six Sigma (and its offshoot, Lean Six Sigma) apply the principles of science to business, teaching managers to methodically target waste, maximize output and streamline the flow from producer to consumer. […]