生成对抗网络GANs学习路线

2019 年 6 月 10 日 专知

导读

MLmindset作者发布了一篇生成对抗网络合集文章,整合了各类关于GAN的资源,如GANs文章、模型、代码、应用、课程、书籍、教程等等。这份完整的GANs资源合集为研究者提供了完美的生成对抗网络学习路线。


https://machinelearningmindset.com/generative-adversarial-networks-roadmap/

作者 | Amirsina Torfi

编译 | Xiaowen


目录

① 简介

② 论文-类型及模型、应用

VanillaGAN、CGAN、LAPGAN、DCGAN、AAE、GRAN、InfoGan、BiGan等

③ 课程

④书籍

⑤ 教程


简介



生成对抗网络Generative Adversarial Network,简称GAN)是非监督式学习的一种方法,通过让两个神经网络相互博弈的方式进行学习。


生成对抗网络由一个生成网络与一个判别网络组成。生成网络从潜在空间(latent space)中随机采样作为输入,其输出结果需要尽量模仿训练集中的真实样本。判别网络的输入则为真实样本或生成网络的输出,其目的是将生成网络的输出从真实样本中尽可能分辨出来。而生成网络则要尽可能地欺骗判别网络。两个网络相互对抗、不断调整参数,最终目的是使判别网络无法判断生成网络的输出结果是否真实。

论文


这一章节是关于近年来GAN的相关发表论文。


                           类型及模型



Image by: Rouzbeh Asghari Shirvani


Core: Generative Adversarial Networks (VanillaGAN)

论文1:Generative Adversarial Nets

文章地址:http://papers.nips.cc/paper/5423-generative-adversarial-nets.pdf

代码地址:https://github.com/goodfeli/adversarial


论文2:ENERGY-BASED GENERATIVE ADVERSARIAL NETWORK

文章地址:https://arxiv.org/pdf/1609.03126v2.pdf

代码地址:https://github.com/buriburisuri/ebgan


论文3:Which Training Methods for GANs do Actually Converge

文章地址:https://arxiv.org/pdf/1801.04406.pdf

代码地址:https://github.com/LMescheder/GAN_stability


Conditional Generative Adversarial Networks (CGAN)

论文1:Conditional generative adversarial nets

文章地址:https://arxiv.org/abs/1411.1784

代码地址:https://github.com/zhangqianhui/Conditional-GAN


论文2:Photo-realistic single image super-resolution using a GAN

文章地址:https://arxiv.org/pdf/1609.04802.pdf

代码地址:https://github.com/tensorlayer/srgan


论文3:Image-to-Image Translation with Conditional Adversarial Networks

文章地址:https://arxiv.org/abs/1611.07004

代码地址:https://github.com/phillipi/pix2pix


论文4:Generative Visual Manipulation on the Natural Image Manifold

文章地址:https://arxiv.org/abs/1609.03552

代码地址:https://github.com/junyanz/iGAN


Laplacian Pyramid of Adversarial Networks (LAPGAN)

论文1:Deep Generative Image Models using a Laplacian Pyramid of Adversarial Networks

文章地址:http://papers.nips.cc/paper/5773-deep-generative-image-models-using-a-laplacian-pyramid-of-adversarial-networks.pdf

代码地址:https://github.com/witnessai/LAPGAN


Deep Convolutional Generative Adversarial Networks (DCGAN)

论文1:Deep Convolutional Generative Adversarial Networks

文章地址:http://papers.nips.cc/paper/5773-deep-generative-image-models-using-a-laplacian-pyramid-of-adversarial-networks.pdf

代码地址:https://github.com/witnessai/LAPGAN


论文2:Generative Adversarial Text to Image Synthesis

文章地址:https://arxiv.org/pdf/1605.05396.pdf

代码地址:https://github.com/reedscot/icml2016


Adversarial Autoencoders (AAE)

论文1:Adversarial Autoencoders

文章地址:https://arxiv.org/abs/1511.05644

代码地址:https://github.com/Naresh1318/Adversarial_Autoencoder


Generative Recurrent Adversarial Networks (GRAN)

论文1:Generating images with recurrent adversarial networks

文章地址:https://arxiv.org/abs/1602.05110

代码地址:https://github.com/jiwoongim/GRAN


Information Maximizing Generative Adversarial Networks (InfoGan)

论文1:Infogan: Information maximizing GANs

文章地址:http://papers.nips.cc/paper/6399-infogan-interpretable-representation

代码地址:https://github.com/openai/InfoGAN


Bidirectional Generative Adversarial Networks (BiGan)

论文1:Adversarial feature learning

文章地址:https://arxiv.org/abs/1605.09782

代码地址:https://github.com/jeffdonahue/bigan


                            应用

GANs理论和训练 Theory and Tranining

论文1:Energy-based generative adversarial network

文章地址:https://arxiv.org/pdf/1609.03126v2.pdf

代码地址:https://github.com/buriburisuri/ebgan


论文2:Which Training Methods for GANs do actually Converge

文章地址:https://arxiv.org/pdf/1801.04406.pdf

代码地址:https://github.com/LMescheder/GAN_stability


论文3:Improved Techniques for Training GANs

文章地址:https://arxiv.org/abs/1609.04468

代码地址:https://github.com/openai/improved-gan


论文4:Towards Principled Methods for Training Generative Adversarial Networks

文章地址:https://arxiv.org/abs/1701.04862


论文5:Least Squares Generative Adversarial Networks

文章地址:https://arxiv.org/abs/1611.04076

代码地址:https://github.com/pfnet-research/chainer-LSGAN


论文6:Wasserstein GAN

文章地址:https://arxiv.org/abs/1701.07875

代码地址:https://github.com/martinarjovsky/WassersteinGAN


论文7:Improved Training of Wasserstein GANs

文章地址:https://arxiv.org/abs/1704.00028

代码地址:https://github.com/igul222/improved_wgan_training


论文8:Generalization and Equilibrium in Generative Adversarial Nets

文章地址:https://arxiv.org/abs/1703.00573


论文9:GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium

文章地址:https://arxiv.org/abs/1605.09782

代码地址:https://github.com/bioinf-jku/TTUR


论文10:Spectral Normalization for Generative Adversarial Networks

文章地址:https://openreview.net/forum?id=B1QRgziT-

代码地址:https://github.com/minhnhat93/tf-SNDCGAN


图像合成 Image Synthesis

论文1:Generative Adversarial Text to Image Synthesis

文章地址:https://arxiv.org/abs/1605.05396

代码地址:https://github.com/reedscot/icml201


论文2:Plug & Play Generative Networks: Conditional Iterative Generation of Images in Latent Space

文章地址:https://arxiv.org/abs/1612.00005v1

代码地址:https://github.com/Evolving-AI-Lab/ppgn


论文3:Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks

文章地址:https://arxiv.org/abs/1511.06434

代码地址:https://github.com/jacobgil/keras-dcgan


论文4:Progressive Growing of GANs for Improved Quality, Stability, and Variation

文章地址:http://research.nvidia.com/publication/2017-10_Progressive-Growing-of

代码地址:https://github.com/tkarras/progressive_growing_of_gans


论文5:StackGAN: Text to Photo-realistic Image Synthesis with Stacked Generative Adversarial Networks

文章地址:https://arxiv.org/pdf/1612.03242v1.pdf

代码地址:https://github.com/hanzhanggit/StackGAN


论文6:Self-Attention Generative Adversarial Networks

文章地址:https://arxiv.org/abs/1805.08318

代码地址:https://github.com/heykeetae/Self-Attention-GAN


论文7:Large Scale GAN Training for High Fidelity Natural Image Synthesis

文章地址:https://arxiv.org/abs/1809.11096


图像翻译 Image-to-image translation

论文1:Image-to-image translation using conditional adversarial nets

文章地址:https://arxiv.org/pdf/1611.07004v1.pdf

代码地址:https://github.com/phillipi/pix2pix


论文2:Learning to Discover Cross-Domain Relations with Generative Adversarial Networks

文章地址:https://arxiv.org/abs/1703.05192

代码地址:https://github.com/carpedm20/DiscoGAN-pytorch


论文3:Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks

文章地址:https://junyanz.github.io/CycleGAN/

代码地址:https://github.com/junyanz/CycleGAN


论文4:CoGAN: Coupled Generative Adversarial Networks

文章地址:https://arxiv.org/abs/1606.07536

代码地址:https://github.com/andrewliao11/CoGAN-tensorflow


论文5:Unsupervised Image-to-Image Translation Networks

文章地址:https://arxiv.org/abs/1703.00848


论文6:High-Resolution Image Synthesis and Semantic Manipulation with Conditional GANs

文章地址:https://arxiv.org/abs/1711.11585


论文7:UNIT: UNsupervised Image-to-image Translation Networks

文章地址:https://arxiv.org/abs/1703.00848

代码地址:https://github.com/mingyuliutw/UNIT


论文8:Multimodal Unsupervised Image-to-Image Translation

文章地址:https://arxiv.org/abs/1804.04732

代码地址:https://github.com/nvlabs/MUNIt


超像素 Super-resolution

论文1:Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network

文章地址:https://arxiv.org/abs/1609.04802

代码地址:https://github.com/leehomyc/Photo-Realistic-Super-Resoluton


论文2:High-Quality Face Image Super-Resolution Using Conditional Generative Adversarial Networks

文章地址:https://arxiv.org/pdf/1707.00737.pdf


论文3:Analyzing Perception-Distortion Tradeoff using Enhanced Perceptual Super-resolution Network

文章地址:https://arxiv.org/pdf/1811.00344.pdf

代码地址:https://github.com/subeeshvasu/2018_subeesh_epsr_eccvw


文本生成图像 Text to Image

论文1:TAC-GAN – Text Conditioned Auxiliary Classifier Generative Adversarial Network

文章地址:https://arxiv.org/pdf/1703.06412.pdf

代码地址:https://github.com/dashayushman/TAC-GAN


论文2:Generative Adversarial Text to Image Synthesis

文章地址:https://arxiv.org/pdf/1605.05396.pdf

代码地址:https://github.com/paarthneekhara/text-to-imag


论文3:Learning What and Where to Draw

文章地址:http://www.scottreed.info/files/nips2016.pdf

代码地址:https://github.com/reedscot/nips2016


图像编辑 Image Editing

论文1:Invertible Conditional GANs for image editing

文章地址:https://arxiv.org/pdf/1611.06355.pdf

代码地址:https://github.com/Guim3/IcGAN


论文2:Image De-raining Using a Conditional Generative Adversarial Network

文章地址:https://arxiv.org/abs/1701.05957

代码地址:https://github.com/hezhangsprinter/ID-CGAN


ETC

论文1:Generating multi-label discrete patient records using generative adversarial networks

文章地址:https://arxiv.org/abs/1703.06490

代码地址:https://github.com/mp2893/medgan


论文2:Adversarial Generation of Natural Language

文章地址:https://arxiv.org/abs/1705.10929


论文3:Language Generation with Recurrent Generative Adversarial Networks without Pre-training

文章地址:https://arxiv.org/abs/1706.01399

代码地址:https://github.com/amirbar/rnn.wgan


论文4:Adversarial ranking for language generation

文章地址:http://papers.nips.cc/paper/6908-adversarial-ranking-for-language-generation

代码地址:https://github.com/desire2020/RankGAN


论文5:Adversarial Training Methods for Semi-Supervised Text Classification

文章地址:https://arxiv.org/abs/1605.07725

代码地址:https://github.com/aonotas/adversarial_text


课程


  • Deep Learning: GANs and Variational Autoencoders by Udemy: [https://www.udemy.com/deep-learning-gans-and-variational-autoencoders/]

  • Differentiable Inference and Generative Models by the University of Toronto: [http://www.cs.toronto.edu/~duvenaud/courses/csc2541/]

  • Learning Generative Adversarial Networks by Udemy: [https://www.udemy.com/learning-generative-adversarial-networks/]


书籍


  • GANs in Action - Deep learning with Generative Adversarial Networks by manning Publications: [https://www.manning.com/books/gans-in-action]


教程


  • GANs from Scratch 1: A deep introduction. With code in PyTorch and TensorFlow: [https://medium.com/ai-society/gans-from-scratch-1-a-deep-introduction-with-code-in-pytorch-and-tensorflow-cb03cdcdba0f]

  • Keep Calm and train a GAN. Pitfalls and Tips on training Generative Adversarial Networks: [https://medium.com/@utk.is.here/keep-calm-and-train-a-gan-pitfalls-and-tips-on-training-generative-adversarial-networks-edd529764aa9]

  • CVPR 2018 Tutorial on GANs: [https://sites.google.com/view/cvpr2018tutorialongans/]

  • Introductory guide to Generative Adversarial Networks (GANs) and their promise!: [https://www.oreilly.com/learning/generative-adversarial-networks-for-beginners]

  • Generative Adversarial Networks for beginners: [https://www.oreilly.com/learning/generative-adversarial-networks-for-beginners]

  • Understanding and building Generative Adversarial Networks(GANs): [https://becominghuman.ai/understanding-and-building-generative-adversarial-networks-gans-8de7c1dc0e25]


-END-

专 · 知

专知,专业可信的人工智能知识分发,让认知协作更快更好!欢迎登录www.zhuanzhi.ai,注册登录专知,获取更多AI知识资料!

欢迎微信扫一扫加入专知人工智能知识星球群,获取最新AI专业干货知识教程视频资料和与专家交流咨询

请加专知小助手微信(扫一扫如下二维码添加),加入专知人工智能主题群,咨询技术商务合作~

专知《深度学习:算法到实战》课程全部完成!550+位同学在学习,现在报名,限时优惠!网易云课堂人工智能畅销榜首位!

点击“阅读原文”,了解报名专知《深度学习:算法到实战》课程

登录查看更多
36

相关内容

生成对抗网络 (Generative Adversarial Network, GAN) 是一类神经网络,通过轮流训练判别器 (Discriminator) 和生成器 (Generator),令其相互对抗,来从复杂概率分布中采样,例如生成图片、文字、语音等。GAN 最初由 Ian Goodfellow 提出,原论文见 Generative Adversarial Networks

知识荟萃

精品入门和进阶教程、论文和代码整理等

更多

查看相关VIP内容、论文、资讯等
最新《生成式对抗网络》简介,25页ppt
专知会员服务
167+阅读 · 2020年6月28日
【论文】结构GANs,Structured GANs,
专知会员服务
14+阅读 · 2020年1月16日
必读的10篇 CVPR 2019【生成对抗网络】相关论文和代码
专知会员服务
31+阅读 · 2020年1月10日
【CCL 2019】ATT-第19期:生成对抗网络 (邱锡鹏)
专知会员服务
48+阅读 · 2019年11月12日
GANs最新综述论文: 生成式对抗网络及其变种如何有用
专知会员服务
70+阅读 · 2019年10月19日
生成式对抗网络GAN异常检测
专知会员服务
114+阅读 · 2019年10月13日
GAN学习路线图:论文、应用、课程、书籍大总结
全球人工智能
16+阅读 · 2019年7月8日
万字综述之生成对抗网络(GAN)
PaperWeekly
43+阅读 · 2019年3月19日
必读!生成对抗网络GAN论文TOP 10
全球人工智能
6+阅读 · 2019年3月19日
若干生成对抗网络模型简介
统计学习与视觉计算组
9+阅读 · 2018年6月13日
历史最全GAN网络及其各种变体整理(附论文及代码实现)
一文读懂生成对抗网络GANs(附学习资源)
数据派THU
10+阅读 · 2018年2月9日
GAN | GAN介绍(2)
中国科学院网络数据重点实验室
43+阅读 · 2017年8月4日
Meta-Transfer Learning for Zero-Shot Super-Resolution
Arxiv
43+阅读 · 2020年2月27日
Seeing What a GAN Cannot Generate
Arxiv
7+阅读 · 2019年10月24日
Arxiv
8+阅读 · 2019年2月15日
Arxiv
7+阅读 · 2018年5月21日
Arxiv
10+阅读 · 2018年3月23日
Arxiv
4+阅读 · 2018年3月23日
VIP会员
相关资讯
GAN学习路线图:论文、应用、课程、书籍大总结
全球人工智能
16+阅读 · 2019年7月8日
万字综述之生成对抗网络(GAN)
PaperWeekly
43+阅读 · 2019年3月19日
必读!生成对抗网络GAN论文TOP 10
全球人工智能
6+阅读 · 2019年3月19日
若干生成对抗网络模型简介
统计学习与视觉计算组
9+阅读 · 2018年6月13日
历史最全GAN网络及其各种变体整理(附论文及代码实现)
一文读懂生成对抗网络GANs(附学习资源)
数据派THU
10+阅读 · 2018年2月9日
GAN | GAN介绍(2)
中国科学院网络数据重点实验室
43+阅读 · 2017年8月4日
Top
微信扫码咨询专知VIP会员