Unsupervised Image-to-Image Translation Networks --- Reading Writing

2017.03.03

  Motivations: most existing image to image translation algorithms are all need image pairs as training data for deep neural network, such as CGANs or VAEs. But in some cases, it is rather difficult to collect such training data. For example, the night and day image pairs, the perfect aligned thermal RGB image pairs, or sunning rainning, fogging, et al, which provide us a new challenging problem:

  How to do image to image translation in a unsupervised fashion which do not need aligned image pairs ?

This paper proposed the UNIT framework (UNsupervised Image-to-image Translation network) to deal with this problem which combine VAE and GANs. The whole framework can be described as the following figures which seems complex but rather easy to understand.

  

  There are two most important assumptions about the proposed framework:

  1. we assume that the relationship between X1 and X2 does not only exist at the image level but also at the level of local patches or regions.

  2. for any given images x1 and x2,  there exists a common underlying representation z, such that we can cover both images from this underlying representation from each of the two input images.

  VAEs: the encoder-generator pair {E1, G1} constitutes a VAE for the X1 domain, termed VAE1. Another pair of {E2, G2} constitutes a VAE for the X2 domain VAE2.

    Weight-sharing : we enforce a weight-sharing constraint to relate the representations in the two VAEs.

  GANs :  two GANs are used to output the two domains.

  

  

  

  


    Experiments:

  

  

    

  

  

  

Unsupervised Image-to-Image Translation Networks --- Reading Writing的更多相关文章

  1. Learning to Compare Image Patches via Convolutional Neural Networks --- Reading Summary

    Learning to Compare Image Patches via Convolutional Neural Networks ---  Reading Summary 2017.03.08 ...

  2. Unsupervised Image-to-Image Translation Networks

    Abstract: 无监督图像到图像的翻译目的是学习不同域图像的一个联合分布,通过使用来自单独域图像的边缘分布.给定一个边缘分布,可以得到很多种联合分布.如果不加入额外的假设条件的话,从边缘分布无法推 ...

  3. reading/writing files in Python

    file types: plaintext files, such as .txt .py Binary files, such as .docx, .pdf, iamges, spreadsheet ...

  4. 【Deep Learning】Hinton. Reducing the Dimensionality of Data with Neural Networks Reading Note

    2006年,机器学习泰斗.多伦多大学计算机系教授Geoffery Hinton在Science发表文章,提出基于深度信念网络(Deep Belief Networks, DBN)可使用非监督的逐层贪心 ...

  5. On Explainability of Deep Neural Networks

    On Explainability of Deep Neural Networks « Learning F# Functional Data Structures and Algorithms is ...

  6. The Unreasonable Effectiveness of Recurrent Neural Networks (RNN)

    http://karpathy.github.io/2015/05/21/rnn-effectiveness/ There’s something magical about Recurrent Ne ...

  7. 26 THINGS I LEARNED IN THE DEEP LEARNING SUMMER SCHOOL

    26 THINGS I LEARNED IN THE DEEP LEARNING SUMMER SCHOOL In the beginning of August I got the chance t ...

  8. DotNet 资源大全中文版(Awesome最新版)

    Awesome系列的.Net资源整理.awesome-dotnet是由quozd发起和维护.内容包括:编译器.压缩.应用框架.应用模板.加密.数据库.反编译.IDE.日志.风格指南等. 算法与数据结构 ...

  9. (转) Awesome - Most Cited Deep Learning Papers

    转自:https://github.com/terryum/awesome-deep-learning-papers Awesome - Most Cited Deep Learning Papers ...

随机推荐

  1. html5-增强的表单

    <!DOCTYPE html><html lang="en"><head>    <meta charset="UTF-8&qu ...

  2. mybatis oracle -批量插入,存在则更新

    <insert id="batchUpdatePBWUserInfo" parameterType="java.util.List"> MERGE ...

  3. GCD(III)

    GCD 线程间的通信 在iOS开发过程中,我们一般在主线程里边进行UI刷新,例如:点击.滚动.拖拽等事件.我们通常把一些耗时的操作放在其他线程,比如说图片下载.文件上传等耗时操作.而当我们有时候在其他 ...

  4. windows下配置lua环境

    1.进入lua官网http://www.lua.org/ 2.点击download 3.点击get a binary 4.点击[Lua - joedf's Builds] 5.选择适合自己的版本下载, ...

  5. 获取 web 服务器 port

    Tomcat: public static String getServerPort(boolean secure) throws AttributeNotFoundException, Instan ...

  6. render函数

    vue2.0之render函数   虽然vue推荐用template来创建你的html,但是在某些时候你也会用到render函数. 虚拟DOM Vue 通过建立一个虚拟 DOM 对真实 DOM 发生的 ...

  7. 【javascript】对原型对象、原型链的理解

    原型对象,原型链这些知识属于基础类知识.但是平时开发过程中也很少用到. 看网上的意思,原型链用于es5开发场景下的继承.es6有了类语法糖之后,就自带继承了. 通过理解,个人画了一张原型链解构的关系图 ...

  8. Oracle之现有表上建新表、操作符、字符函数

    #PLSQL技术培训15页PPT利用现有表创建表(百度) 说明:做新操作前要对旧表备份  具体百度 语法: create table <new_table_name> as select ...

  9. ELK学习笔记之F5利用EELK进行应用数据挖掘系列(2)-DNS

    0x00 概述 很多客户使用GTM/DNS为企业业务提供动态智能解析,解决应用就近性访问.优选问题.对于已经实施多数据中心双活的客户,则会使用GSLB提供双活流量调度.DNS作为企业业务访问的指路者, ...

  10. Python查看关键字和帮助信息

    1.查看所有的关键字 >>> help('keywords') Here is a list of the Python keywords. Enter any keyword to ...