{"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/xunit-learning-a-spatial-activation-function","title":"xUnit: Learning a Spatial Activation Function for Efficient Image Restoration","arxiv_id":"1711.06445","date":"2017-11-17","proceeding":"CVPR 2018 6","authors":["Idan Kligvasser","Tamar Rott Shaham","Tomer Michaeli"],"abstract":"In recent years, deep neural networks (DNNs) achieved unprecedented\nperformance in many low-level vision tasks. However, state-of-the-art results\nare typically achieved by very deep networks, which can reach tens of layers\nwith tens of millions of parameters. To make DNNs implementable on platforms\nwith limited resources, it is necessary to weaken the tradeoff between\nperformance and efficiency. In this paper, we propose a new activation unit,\nwhich is particularly suitable for image restoration problems. In contrast to\nthe widespread per-pixel activation units, like ReLUs and sigmoids, our unit\nimplements a learnable nonlinear function with spatial connections. This\nenables the net to capture much more complex features, thus requiring a\nsignificantly smaller number of layers in order to reach the same performance.\nWe illustrate the effectiveness of our units through experiments with\nstate-of-the-art nets for denoising, de-raining, and super resolution, which\nare already considered to be very small. With our approach, we are able to\nfurther reduce these models by nearly 50% without incurring any degradation in\nperformance.","url_abs":"http://arxiv.org/abs/1711.06445v3","url_pdf":"http://arxiv.org/pdf/1711.06445v3.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":"xunit-learning-a-spatial-activation-function","repo_url":"https://github.com/kligvasser/xUnit","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"denoising","task_name":"Denoising"},{"task_slug":"image-restoration","task_name":"Image Restoration"},{"task_slug":"super-resolution","task_name":"Super-Resolution"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1711.06445","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}