跳至主要內容

GAN

Vingkin...大约 2 分钟

DG(x)=pdata (x)pdata (x)+pg(x)D_{G}^{*}(\boldsymbol{x})=\frac{p_{\text {data }}(\boldsymbol{x})}{p_{\text {data }}(\boldsymbol{x})+p_{g}(\boldsymbol{x})}

附录

生成对抗网络GAN开山之作

生成对抗神经网络GAN开山之作论文:Generative Adversarial Nets。

论文作者为“生成对抗网络之父”Ian Goodfellow和图灵奖得主Youshua Bengio。

GAN近年来成为人工智能和深度学习的热门研究领域。GAN广泛应用于图像生成、风格迁移、AI艺术、黑白老照片上色修复。你可以使用GAN实现照片转成油画、野马转成斑马、黑夜转成白天,简笔画的猫转成真猫,模糊图像转成高清图像等酷炫好玩的应用。

子豪兄会带你逐句精读GAN开山之作经典论文,讲解GAN精巧的自监督对抗学习范式背后的算法和数学原理。

OpenMMLab图像生成开源算法库MMGeneration:https://github.com/open-mmlab/mmgenerationopen in new window

原始论文:https://proceedings.neurips.cc/paper/2014/file/5ca3e9b122f61f8f06494c97b1afccf3-Paper.pdfopen in new window

论文主页:https://proceedings.neurips.cc/paper/2014/hash/5ca3e9b122f61f8f06494c97b1afccf3-Abstract.htmlopen in new window

趣味demo

交互式可视化GAN Lab:https://poloclub.github.io/ganlabopen in new window

交互式可视化GAN Lab视频讲解:https://www.bilibili.com/video/BV1R44y1377Topen in new window

照片转动漫AnimeGAN:https://github.com/TachibanaYoshino/AnimeGANopen in new window

英伟达GauGAN2神笔马良:http://gaugan.org/gaugan2/open in new window

不存在的人脸:https://thispersondoesnotexist.com/open in new window

不存在的花瓶:https://thisvesseldoesnotexist.comopen in new window

不存在的梗图:https://imgflip.com/ai-memeopen in new window

黑白老照片上色:https://deepai.org/machine-learning-model/colorizeropen in new window

黑白老照片上色:https://github.com/jantic/DeOldifyopen in new window

黑白老照片上色:https://www.myheritage.com/incoloropen in new window

OpenAI DALLE-2:https://openai.com/dall-e-2open in new window

作者相关

Schmidhuber在2016年NIPS大会上打断Goodfellow的GAN Tutorial:https://zhuanlan.zhihu.com/p/27159510open in new window

【AI大咖】扒一下低调的Yoshua Bengio大神

https://zhuanlan.zhihu.com/p/66259338open in new window

博客

李宏毅生成对抗网络公开课:https://www.bilibili.com/video/av24011528/open in new window

理解生成对抗网络:https://danieltakeshi.github.io/2017/03/05/understanding-generative-adversarial-networks/open in new window

GAN动物园:https://github.com/hindupuravinash/the-gan-zooopen in new window

代码复现

官方论文复现:https://github.com/goodfeli/adversarialopen in new window

OpenMMLab开源图像生成算法库MMGeneration:https://github.com/open-mmlab/mmgenerationopen in new window

各种GAN的Keras实现:https://github.com/eriklindernoren/Keras-GANopen in new window

GAN的推导及Keras手把手实现(老版本Keras):http://srome.github.io/An-Annotated-Proof-of-Generative-Adversarial-Networks-with-Implementation-Notes/open in new window

评论
  • 按正序
  • 按倒序
  • 按热度
Powered by Waline v2.15.8