ECCV2018 超分辨相关.7z

  • v6_564993
    了解作者
  • 6.9MB
    文件大小
  • 7z
    文件格式
  • 0
    收藏次数
  • VIP专享
    资源类型
  • 0
    下载次数
  • 2022-05-19 08:34
    上传日期
亲历了ECCV 2018的机器学习研究员Tetianka Martyniuk挑选了6篇ECCV 2018接收论文,概述了超分辨率(Super-Resolution, SR)技术的未来发展趋势。内容包括:一:学习图像超分辨率,先学习图像退化;二:由面部五官热力图指导的面部超分辨率;三:用深度残差通道的注意网络的图像超分辨率;四:用于图像超分辨率的多尺度残差网络;五:级联残差加持的快速、准确、轻量级的超分辨率网络;六:SRFeat:具有特征识别的单个图像超分辨率
ECCV2018 超分辨相关.7z
  • ECCV2018 超分辨相关
  • Multi-scale Residual Network for Image Super-Resolution.pdf
    1.6MB
  • Fast, Accurate, and Lightweight Super-Resolution with Cascading Residual Network.pdf
    942.1KB
  • Image Super-Resolution Using Very Deep Residual Channel Attention Networks.pdf
    862.3KB
  • Face Super-resolution Guided by Facial Component Heatmaps.pdf
    784.3KB
  • Single Image Super-Resolution with Feature Discrimination.pdf
    1.2MB
  • To learn image super-resolution, use a GAN to learn how to do image degradation first.pdf
    3.1MB
内容介绍
<html xmlns="http://www.w3.org/1999/xhtml"> <head> <meta charset="utf-8"> <meta name="generator" content="pdf2htmlEX"> <meta http-equiv="X-UA-Compatible" content="IE=edge,chrome=1"> <link rel="stylesheet" href="https://static.pudn.com/base/css/base.min.css"> <link rel="stylesheet" href="https://static.pudn.com/base/css/fancy.min.css"> <link rel="stylesheet" href="https://static.pudn.com/prod/directory_preview_static/628590921d352a14fcd54d3a/raw.css"> <script src="https://static.pudn.com/base/js/compatibility.min.js"></script> <script src="https://static.pudn.com/base/js/pdf2htmlEX.min.js"></script> <script> try{ pdf2htmlEX.defaultViewer = new pdf2htmlEX.Viewer({}); }catch(e){} </script> <title></title> </head> <body> <div id="sidebar" style="display: none"> <div id="outline"> </div> </div> <div id="pf1" class="pf w0 h0" data-page-no="1"><div class="pc pc1 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://static.pudn.com/prod/directory_preview_static/628590921d352a14fcd54d3a/bg1.jpg"><div class="t m0 x1 h2 y1 ff1 fs0 fc0 sc0 ls0 ws0">T<span class="_ _0"></span>o<span class="_ _1"> </span>learn<span class="_ _1"> </span>image<span class="_ _1"> </span>sup<span class="_ _2"></span>er-resolution,<span class="_ _1"> </span>use<span class="_ _1"> </span>a<span class="_ _1"> </span>GAN<span class="_ _1"> </span>to</div><div class="t m0 x2 h2 y2 ff1 fs0 fc0 sc0 ls0 ws0">learn<span class="_ _1"> </span>ho<span class="_ _3"></span>w<span class="_ _1"> </span>to<span class="_ _1"> </span>do<span class="_ _1"> </span>image<span class="_ _1"> </span>degradation<span class="_ _1"> </span>&#64257;rst</div><div class="t m0 x3 h3 y3 ff2 fs1 fc0 sc0 ls0 ws0">Adrian<span class="_ _4"> </span>Bulat*,<span class="_ _4"> </span>Jing<span class="_ _4"> </span>Y<span class="_ _5"></span>ang*,<span class="_ _4"> </span>Georgios<span class="_ _4"> </span>Tzimirop<span class="_ _2"></span>oulos</div><div class="t m0 x4 h4 y4 ff3 fs2 fc0 sc0 ls0 ws0">Computer<span class="_ _4"> </span>Vision<span class="_ _6"> </span>Lab<span class="_ _2"></span>oratory<span class="_ _5"></span>,<span class="_ _6"> </span>Universit<span class="_ _3"></span>y<span class="_ _6"> </span>of<span class="_ _4"> </span>Nottingham,<span class="_ _6"> </span>U.K.</div><div class="t m0 x5 h4 y5 ff4 fs2 fc0 sc0 ls0 ws0">{<span class="ff5">adrian.bulat,jing.yang2,yorgos.tzimiropoulos</span>}<span class="ff5">@nottingham.ac.uk</span></div><div class="t m0 x6 h4 y6 ff6 fs2 fc0 sc0 ls0 ws0">Abstract.<span class="_ _7"> </span><span class="ff3">This<span class="_ _4"> </span>pap<span class="_ _2"></span>er<span class="_ _4"> </span>is<span class="_ _4"> </span>on<span class="_ _4"> </span>image<span class="_ _8"> </span>and<span class="_ _4"> </span>face<span class="_ _4"> </span>sup<span class="_ _2"></span>er-resolution.<span class="_ _4"> </span>The<span class="_ _4"> </span>v<span class="_ _3"></span>ast</span></div><div class="t m0 x6 h4 y7 ff3 fs2 fc0 sc0 ls0 ws0">ma<span class="_ _2"></span>jority<span class="_ _9"> </span>of<span class="_ _9"> </span>prior<span class="_ _a"> </span>work<span class="_ _9"> </span>for<span class="_ _9"> </span>this<span class="_ _a"> </span>problem<span class="_ _9"> </span>fo<span class="_ _2"></span>cus<span class="_ _9"> </span>on<span class="_ _a"> </span>how<span class="_ _9"> </span>to<span class="_ _9"> </span>increase<span class="_ _a"> </span>the</div><div class="t m0 x6 h4 y8 ff3 fs2 fc0 sc0 ls0 ws0">resolution<span class="_ _a"> </span>of<span class="_ _9"> </span>low-resolution<span class="_ _9"> </span>imag<span class="_"> </span>es<span class="_ _a"> </span>which<span class="_ _9"> </span>are<span class="_ _a"> </span>arti&#64257;cially<span class="_ _9"> </span>g<span class="_"> </span>en<span class="_"> </span>era<span class="_"> </span>t<span class="_"> </span>ed<span class="_ _a"> </span>by</div><div class="t m0 x6 h4 y9 ff3 fs2 fc0 sc0 ls0 ws0">simple<span class="_ _6"> </span>bilinear<span class="_ _6"> </span>do<span class="_ _3"></span>wn-sampling<span class="_ _6"> </span>(or<span class="_ _6"> </span>in<span class="_ _6"> </span>a<span class="_ _6"> </span>few<span class="_ _6"> </span>cases<span class="_ _6"> </span>b<span class="_ _3"></span>y<span class="_ _6"> </span>blurring<span class="_ _6"> </span>follow<span class="_ _3"></span>ed<span class="_ _6"> </span>b<span class="_ _3"></span>y</div><div class="t m0 x6 h4 ya ff3 fs2 fc0 sc0 ls0 ws0">down-sampling).<span class="_ _b"> </span>W<span class="_ _5"></span>e<span class="_ _b"> </span>show<span class="_ _b"> </span>that<span class="_ _c"> </span>such<span class="_ _b"> </span>methods<span class="_ _c"> </span>fail<span class="_ _b"> </span>to<span class="_ _c"> </span>pro<span class="_ _2"></span>duce<span class="_ _b"> </span>go<span class="_ _2"></span>od<span class="_ _b"> </span>results</div><div class="t m0 x6 h4 yb ff3 fs2 fc0 sc0 ls0 ws0">when<span class="_ _b"> </span>applied<span class="_ _b"> </span>to<span class="_ _c"> </span>real-w<span class="_ _3"></span>orld<span class="_ _b"> </span>low-resolution,<span class="_ _b"> </span>low<span class="_ _b"> </span>qualit<span class="_ _3"></span>y<span class="_ _b"> </span>images.<span class="_ _c"> </span>T<span class="_ _5"></span>o<span class="_ _b"> </span>circum-</div><div class="t m0 x6 h4 yc ff3 fs2 fc0 sc0 ls0 ws0">ve<span class="_ _3"></span>nt<span class="_ _6"> </span>this<span class="_ _4"> </span>problem,<span class="_ _6"> </span>we<span class="_ _6"> </span>prop<span class="_ _2"></span>ose<span class="_ _6"> </span>a<span class="_ _4"> </span>t<span class="_ _3"></span>wo-stage<span class="_ _6"> </span>pro<span class="_ _2"></span>cess<span class="_ _6"> </span>which<span class="_ _6"> </span>&#64257;rstly<span class="_ _4"> </span>trains<span class="_ _4"> </span>a</div><div class="t m0 x6 h4 yd ff3 fs2 fc0 sc0 ls0 ws0">High-to-Low<span class="_ _b"> </span>Generative<span class="_ _c"> </span>Adv<span class="_ _3"></span>ersarial<span class="_ _c"> </span>Netw<span class="_ _3"></span>ork<span class="_ _c"> </span>(GAN)<span class="_ _c"> </span>to<span class="_ _c"> </span>learn<span class="_ _c"> </span>how<span class="_ _c"> </span>to<span class="_ _c"> </span>de-</div><div class="t m0 x6 h4 ye ff3 fs2 fc0 sc0 ls0 ws0">grade<span class="_ _c"> </span>and<span class="_ _c"> </span>do<span class="_ _3"></span>wnsample<span class="_ _c"> </span>high-resolution<span class="_ _c"> </span>images<span class="_ _c"> </span>requiring,<span class="_ _c"> </span>during<span class="_ _c"> </span>training,</div><div class="t m0 x6 h4 yf ff3 fs2 fc0 sc0 ls0 ws0">only<span class="_ _6"> </span><span class="ff7">unp<span class="_ _3"></span>air<span class="_ _3"></span>e<span class="_ _3"></span>d<span class="_ _4"> </span><span class="ff3">hig<span class="_"> </span>h<span class="_ _6"> </span>and<span class="_ _6"> </span>low-resolution<span class="_ _6"> </span>images.<span class="_ _6"> </span>Once<span class="_ _c"> </span>t<span class="_"> </span>h<span class="_"> </span>is<span class="_ _6"> </span>is<span class="_ _6"> </span>achiev<span class="_ _3"></span>ed,<span class="_ _6"> </span>the</span></span></div><div class="t m0 x6 h4 y10 ff3 fs2 fc0 sc0 ls0 ws0">output<span class="_ _8"> </span>of<span class="_ _9"> </span>this<span class="_ _8"> </span>netw<span class="_ _3"></span>ork<span class="_ _9"> </span>is<span class="_ _8"> </span>used<span class="_ _9"> </span>to<span class="_ _8"> </span>train<span class="_ _8"> </span>a<span class="_ _9"> </span>Low-to-High<span class="_ _8"> </span>GAN<span class="_ _8"> </span>for<span class="_ _9"> </span>image</div><div class="t m0 x6 h4 y11 ff3 fs2 fc0 sc0 ls0 ws0">sup<span class="_ _2"></span>er-resolution<span class="_ _6"> </span>using<span class="_ _4"> </span>this<span class="_ _6"> </span>time<span class="_ _4"> </span><span class="ff7">p<span class="_ _3"></span>air<span class="_ _3"></span>e<span class="_ _3"></span>d<span class="_ _9"> </span><span class="ff3">lo<span class="_ _3"></span>w-<span class="_ _4"> </span>and<span class="_ _6"> </span>high-<span class="_"> </span>res<span class="_"> </span>ol<span class="_"> </span>u<span class="_"> </span>t<span class="_"> </span>io<span class="_"> </span>n<span class="_ _4"> </span>images.</span></span></div><div class="t m0 x6 h4 y12 ff3 fs2 fc0 sc0 ls0 ws0">Our<span class="_ _4"> </span>main<span class="_ _4"> </span>result<span class="_ _8"> </span>is<span class="_ _4"> </span>that<span class="_ _4"> </span>this<span class="_ _8"> </span>net<span class="_ _3"></span>work<span class="_ _4"> </span>can<span class="_ _4"> </span>b<span class="_ _2"></span>e<span class="_ _4"> </span>no<span class="_ _3"></span>w<span class="_ _4"> </span>used<span class="_ _4"> </span>to<span class="_ _8"> </span>e&#64256;ectively<span class="_ _4"> </span>in-</div><div class="t m0 x6 h4 y13 ff3 fs2 fc0 sc0 ls0 ws0">crease<span class="_ _4"> </span>the<span class="_ _4"> </span>quality<span class="_ _4"> </span>of<span class="_ _4"> </span>real-world<span class="_ _4"> </span>lo<span class="_ _3"></span>w-resolution<span class="_ _4"> </span>images.<span class="_ _8"> </span>W<span class="_ _5"></span>e<span class="_ _4"> </span>hav<span class="_ _3"></span>e<span class="_ _4"> </span>applied</div><div class="t m0 x6 h4 y14 ff3 fs2 fc0 sc0 ls0 ws0">the<span class="_ _6"> </span>prop<span class="_ _2"></span>osed<span class="_ _6"> </span>pip<span class="_ _2"></span>eline<span class="_ _6"> </span>for<span class="_ _4"> </span>the<span class="_ _6"> </span>problem<span class="_ _4"> </span>of<span class="_ _6"> </span>face<span class="_ _6"> </span>sup<span class="_ _2"></span>er-resolution<span class="_ _6"> </span>where<span class="_ _4"> </span>w<span class="_ _3"></span>e</div><div class="t m0 x6 h4 y15 ff3 fs2 fc0 sc0 ls0 ws0">rep<span class="_ _2"></span>ort<span class="_ _9"> </span>large<span class="_ _a"> </span>impro<span class="_ _3"></span>vemen<span class="_ _3"></span>t<span class="_ _a"> </span>o<span class="_ _3"></span>ver<span class="_ _9"> </span>baselines<span class="_ _a"> </span>and<span class="_ _a"> </span>prior<span class="_ _9"> </span>work<span class="_ _a"> </span>although<span class="_ _9"> </span>the</div><div class="t m0 x6 h4 y16 ff3 fs2 fc0 sc0 ls0 ws0">prop<span class="_ _2"></span>osed<span class="_ _6"> </span>metho<span class="_ _2"></span>d<span class="_ _6"> </span>is<span class="_ _6"> </span>p<span class="_ _2"></span>oten<span class="_ _3"></span>tially<span class="_ _4"> </span>applicable<span class="_ _6"> </span>to<span class="_ _4"> </span>other<span class="_ _6"> </span>ob<span class="_ _2"></span>ject<span class="_ _4"> </span>categories.</div><div class="t m0 x6 h4 y17 ff6 fs2 fc0 sc0 ls0 ws0">Keyw<span class="_ _3"></span>ords:<span class="_ _7"> </span><span class="ff3">Image<span class="_ _a"> </span>an<span class="_"> </span>d<span class="_ _7"> </span>face<span class="_ _7"> </span>sup<span class="_ _2"></span>er-resolution,<span class="_ _d"> </span>Generative<span class="_ _a"> </span>Adversarial</span></div><div class="t m0 x6 h4 y18 ff3 fs2 fc0 sc0 ls0 ws0">Netw<span class="_ _3"></span>orks,<span class="_ _6"> </span>GANs.</div><div class="t m0 x7 h5 y19 ff1 fs3 fc0 sc0 ls0 ws0">1<span class="_ _e"> </span>In<span class="_ _3"></span>tro<span class="_ _2"></span>duction</div><div class="t m0 x8 h3 y1a ff2 fs1 fc0 sc0 ls0 ws0">This<span class="_ _4"> </span>pap<span class="_ _2"></span>er<span class="_ _4"> </span>is<span class="_ _4"> </span>on<span class="_ _8"> </span>enhancing<span class="_ _4"> </span>the<span class="_ _8"> </span>resolution<span class="_ _8"> </span>and<span class="_ _4"> </span>quality<span class="_ _4"> </span>of<span class="_ _4"> </span>low-resolution,<span class="_ _4"> </span>noisy<span class="_ _5"></span>,</div><div class="t m0 x7 h3 y1b ff2 fs1 fc0 sc0 ls0 ws0">blurry<span class="_ _5"></span>,<span class="_ _6"> </span>and<span class="_ _6"> </span>corrupted<span class="_ _6"> </span>b<span class="_ _3"></span>y<span class="_ _6"> </span>artefacts<span class="_ _6"> </span>images.<span class="_ _6"> </span>W<span class="_ _5"></span>e<span class="_ _6"> </span>collectiv<span class="_ _3"></span>ely<span class="_ _6"> </span>refer<span class="_ _6"> </span>to<span class="_ _6"> </span>all<span class="_ _6"> </span>these<span class="_ _c"> </span>t<span class="_"> </span>ask<span class="_"> </span>s</div><div class="t m0 x7 h3 y1c ff2 fs1 fc0 sc0 ls0 ws0">as<span class="_ _6"> </span>image<span class="_ _6"> </span>sup<span class="_ _2"></span>er-resolution.<span class="_ _6"> </span>This<span class="_ _6"> </span>is<span class="_ _4"> </span>a<span class="_ _6"> </span>c<span class="_ _3"></span>hallenging<span class="_ _6"> </span>proble<span class="_"> </span>m<span class="_ _4"> </span>with<span class="_ _6"> </span>a<span class="_ _6"> </span>multitude<span class="_ _6"> </span>of<span class="_ _6"> </span>ap-</div><div class="t m0 x7 h3 y1d ff2 fs1 fc0 sc0 ls0 ws0">plications<span class="_ _6"> </span>from<span class="_ _4"> </span>image<span class="_ _6"> </span>enhancement<span class="_ _6"> </span>and<span class="_ _4"> </span>editing<span class="_ _6"> </span>to<span class="_ _6"> </span>image<span class="_ _4"> </span>recognition<span class="_ _6"> </span>and<span class="_ _4"> </span>ob<span class="_ _2"></span>ject</div><div class="t m0 x7 h3 y1e ff2 fs1 fc0 sc0 ls0 ws0">detection<span class="_ _4"> </span>to<span class="_ _4"> </span>name<span class="_ _4"> </span>a<span class="_ _4"> </span>few.</div><div class="t m0 x9 h3 y1f ff2 fs1 fc0 sc0 ls0 ws0">Our<span class="_ _c"> </span>main<span class="_ _6"> </span>fo<span class="_ _2"></span>cus<span class="_ _c"> </span>is<span class="_ _c"> </span>on<span class="_ _6"> </span>the<span class="_ _c"> </span>pr<span class="_"> </span>ob<span class="_"> </span>le<span class="_"> </span>m<span class="_ _6"> </span>of<span class="_ _6"> </span>super-resolving<span class="_ _6"> </span><span class="ff8">r<span class="_ _3"></span>e<span class="_ _3"></span>al-world<span class="_ _6"> </span>low-r<span class="_ _3"></span>esolution</span></div><div class="t m0 x7 h3 y20 ff2 fs1 fc0 sc0 ls0 ws0">images<span class="_ _9"> </span><span class="ff8">for<span class="_ _a"> </span>a<span class="_ _a"> </span>sp<span class="_ _5"></span>e<span class="_ _3"></span>ci&#64257;c<span class="_ _a"> </span>obje<span class="_ _3"></span>ct<span class="_ _9"> </span>c<span class="_ _3"></span>ate<span class="_ _3"></span>gory<span class="ff2">.<span class="_ _9"> </span>W<span class="_ _5"></span>e<span class="_ _a"> </span>use<span class="_ _9"> </span>faces<span class="_ _9"> </span>in<span class="_ _a"> </span>our<span class="_ _9"> </span>case<span class="_ _9"> </span>noting<span class="_ _a"> </span>ho<span class="_ _3"></span>wev<span class="_ _3"></span>er</span></span></div><div class="t m0 x7 h3 y21 ff2 fs1 fc0 sc0 ls0 ws0">that<span class="_ _9"> </span>the<span class="_ _9"> </span>prop<span class="_ _2"></span>osed<span class="_ _9"> </span>metho<span class="_ _2"></span>d<span class="_ _9"> </span>is<span class="_ _9"> </span>p<span class="_ _2"></span>oten<span class="_ _3"></span>tially<span class="_ _9"> </span>applicable<span class="_ _a"> </span>to<span class="_ _9"> </span>other<span class="_ _9"> </span>ob<span class="_ _2"></span>ject<span class="_ _9"> </span>categorie<span class="_"> </span>s.</div><div class="t m0 x7 h3 y22 ff2 fs1 fc0 sc0 ls0 ws0">Although<span class="_ _4"> </span>there<span class="_ _6"> </span>is<span class="_ _4"> </span>a<span class="_ _6"> </span>multitude<span class="_ _6"> </span>of<span class="_ _4"> </span>papers<span class="_ _4"> </span>on<span class="_ _6"> </span>image<span class="_ _4"> </span>and<span class="_ _6"> </span>face<span class="_ _4"> </span>super-resolution,<span class="_ _4"> </span>the</div><div class="t m0 x7 h3 y23 ff2 fs1 fc0 sc0 ls0 ws0">large<span class="_ _4"> </span>ma<span class="_ _2"></span>jority<span class="_ _4"> </span>of<span class="_ _8"> </span>them<span class="_ _4"> </span>use<span class="_ _8"> </span>as<span class="_ _8"> </span>input<span class="_ _4"> </span>low-resolution<span class="_ _4"> </span>images<span class="_ _8"> </span>which<span class="_ _4"> </span>are<span class="_ _4"> </span>arti&#64257;cially</div><div class="t m0 x7 h3 y24 ff2 fs1 fc0 sc0 ls0 ws0">generated<span class="_ _9"> </span>by<span class="_ _9"> </span>simple<span class="_ _a"> </span>bilinear<span class="_ _9"> </span>down-sampling<span class="_ _9"> </span>or<span class="_ _9"> </span>in<span class="_ _a"> </span>a<span class="_ _9"> </span>few<span class="_ _9"> </span>c<span class="_"> </span>ases<span class="_ _a"> </span>b<span class="_ _3"></span>y<span class="_ _a"> </span>blurring<span class="_ _9"> </span>fol-</div><div class="t m0 x7 h3 y25 ff2 fs1 fc0 sc0 ls0 ws0">lo<span class="_ _3"></span>wed<span class="_ _9"> </span>by<span class="_ _9"> </span>do<span class="_ _3"></span>wn-sampling.<span class="_ _a"> </span>On<span class="_ _9"> </span>the<span class="_ _9"> </span>contrary<span class="_ _5"></span>,<span class="_ _9"> </span>the<span class="_ _a"> </span>real-w<span class="_ _3"></span>orld<span class="_ _9"> </span>se<span class="_"> </span>t<span class="_"> </span>t<span class="_"> </span>in<span class="_"> </span>g<span class="_ _a"> </span>has<span class="_ _9"> </span>received</div><div class="t m0 xa h4 y26 ff3 fs2 fc0 sc0 ls0 ws0">*<span class="_ _6"> </span>Denotes<span class="_ _4"> </span>equal<span class="_ _6"> </span>contribution.</div></div><div class="pi" data-data='{"ctm":[2.206074,0.000000,0.000000,2.206074,0.000000,0.000000]}'></div></div> </body> </html>
评论
    相关推荐
    • 机器学习
      机器学习
    • 机器学习
      机器学习
    • 机器学习
      机器学习 这段代码仅仅是从Web克隆而来的,我们必须使用自定义输入来提高其性能。 首先,使用第一本书的四个基本输入内容。标题2.类型3.作者
    • 机器学习
      机器学习
    • 机器学习
      机器学习
    • 机器学习
      机器学习
    • 机器学习
      机器学习
    • 机器学习
      机器学习 머신러닝연습용코드저장소
    • 机器学习
      机器学习项目 创建了关于宫颈癌预测的ML项目。
    • 机器学习
      机器学习 Используемыедатасеты: label.csv positive.csv,negative.csvсвзятысайта