{"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/memnet-a-persistent-memory-network-for-image","title":"MemNet: A Persistent Memory Network for Image Restoration","arxiv_id":"1708.02209","date":"2017-08-07","proceeding":"ICCV 2017 10","authors":["Ying Tai","Jian Yang","Xiaoming Liu","Chunyan Xu"],"abstract":"Recently, very deep convolutional neural networks (CNNs) have been attracting\nconsiderable attention in image restoration. However, as the depth grows, the\nlong-term dependency problem is rarely realized for these very deep models,\nwhich results in the prior states/layers having little influence on the\nsubsequent ones. Motivated by the fact that human thoughts have persistency, we\npropose a very deep persistent memory network (MemNet) that introduces a memory\nblock, consisting of a recursive unit and a gate unit, to explicitly mine\npersistent memory through an adaptive learning process. The recursive unit\nlearns multi-level representations of the current state under different\nreceptive fields. The representations and the outputs from the previous memory\nblocks are concatenated and sent to the gate unit, which adaptively controls\nhow much of the previous states should be reserved, and decides how much of the\ncurrent state should be stored. We apply MemNet to three image restoration\ntasks, i.e., image denosing, super-resolution and JPEG deblocking.\nComprehensive experiments demonstrate the necessity of the MemNet and its\nunanimous superiority on all three tasks over the state of the arts. Code is\navailable at https://github.com/tyshiwo/MemNet.","url_abs":"http://arxiv.org/abs/1708.02209v1","url_pdf":"http://arxiv.org/pdf/1708.02209v1.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":"memnet-a-persistent-memory-network-for-image","repo_url":"https://github.com/tyshiwo/MemNet","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"memnet-a-persistent-memory-network-for-image","repo_url":"https://github.com/rshwndsz/denoiser","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"color-image-denoising","task_name":"Color Image Denoising"},{"task_slug":"image-restoration","task_name":"Image Restoration"},{"task_slug":"image-super-resolution","task_name":"Image Super-Resolution"},{"task_slug":"jpeg-artifact-correction","task_name":"JPEG Artifact Correction"},{"task_slug":"super-resolution","task_name":"Super-Resolution"}],"methods":[{"method_slug":"memory-network","method_name":"Memory Network"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/color-image-denoising-on-cbsd68-sigma50","task":"Color Image Denoising","dataset":"CBSD68 sigma50","model":"MemNet","rank_in_archive_order":18,"of":18,"metrics":{"PSNR":"26.33"},"uses_additional_data":false},{"leaderboard":"/sota/image-super-resolution-on-bsd100-4x-upscaling","task":"Image Super-Resolution","dataset":"BSD100 - 4x upscaling","model":"MemNet","rank_in_archive_order":45,"of":71,"metrics":{"PSNR":"27.40","SSIM":"0.7281"},"uses_additional_data":false},{"leaderboard":"/sota/image-super-resolution-on-manga109-4x","task":"Image Super-Resolution","dataset":"Manga109 - 4x upscaling","model":"MemNet","rank_in_archive_order":41,"of":50,"metrics":{"PSNR":"29.42","SSIM":"0.8942"},"uses_additional_data":false},{"leaderboard":"/sota/image-super-resolution-on-set14-4x-upscaling","task":"Image Super-Resolution","dataset":"Set14 - 4x upscaling","model":"MemNet","rank_in_archive_order":75,"of":104,"metrics":{"PSNR":"28.26","SSIM":"0.7723"},"uses_additional_data":false},{"leaderboard":"/sota/image-super-resolution-on-urban100-4x","task":"Image Super-Resolution","dataset":"Urban100 - 4x upscaling","model":"MemNet","rank_in_archive_order":51,"of":65,"metrics":{"PSNR":"25.50","SSIM":"0.7630"},"uses_additional_data":false},{"leaderboard":"/sota/jpeg-artifact-correction-on-classic5-quality","task":"JPEG Artifact Correction","dataset":"Classic5 (Quality 10 Grayscale)","model":"MemNet","rank_in_archive_order":6,"of":7,"metrics":{"PSNR":"29.69"},"uses_additional_data":false},{"leaderboard":"/sota/jpeg-artifact-correction-on-live1-quality-10","task":"JPEG Artifact Correction","dataset":"LIVE1 (Quality 10 Color)","model":"MemNet","rank_in_archive_order":7,"of":9,"metrics":{"PSNR":"27.33","PSNR-B":"27.34","SSIM":"0.810"},"uses_additional_data":false},{"leaderboard":"/sota/jpeg-artifact-correction-on-live1-quality-20","task":"JPEG Artifact Correction","dataset":"LIVE1 (Quality 20 Color)","model":"MemNet","rank_in_archive_order":7,"of":9,"metrics":{"PSNR":"29.76","PSNR-B":"29.75","SSIM":"0.877"},"uses_additional_data":false},{"leaderboard":"/sota/jpeg-artifact-correction-on-live1-quality-20-1","task":"JPEG Artifact Correction","dataset":"LIVE1 (Quality 20 Grayscale)","model":"MemNet","rank_in_archive_order":8,"of":12,"metrics":{"PSNR":"31.83","PSNR-B":"31.74","SSIM":"0.8970"},"uses_additional_data":false},{"leaderboard":"/sota/jpeg-artifact-correction-on-live1-quality-10-1","task":"JPEG Artifact Correction","dataset":"Live1 (Quality 10 Grayscale)","model":"MemNet","rank_in_archive_order":7,"of":13,"metrics":{"PSNR":"29.45","PSNR-B":"29.39","SSIM":"0.8327"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1708.02209","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}