经典GANs模型的复现
主要是想通过复现一些简单经典的GANs模型来学习pytorch,之前论文的代码一直用tensorflow,听说工业界pytorch用的居多,最近几年pytorch也比tensorflow更被人看好,因此趁此机会学习一下。
复现的文章有:
(1)最开始的GANs:
题目:Generative Adversarial Nets
文章链接: https://arxiv.org/pdf/1406.2661.pdf .
代码链接: https://github.com/Weifanwong/Implementation-of-classical-GANs-framework/tree/master/GANs.
(2)CGANs:
题目:Conditional Generative Adversarial Nets
文章链接:https://arxiv.org/pdf/1411.1784.pdf.
代码链接:https://github.com/Weifanwong/Implementation-of-classical-GANs-framework/tree/master/CGANs.
(3)WGANs:
题目:Wasserstein GAN
文章链接:https://arxiv.org/pdf/1701.07875.pdf.
代码链接:https://github.com/Weifanwong/Implementation-of-classical-GANs-framework/tree/master/WGANs.
(4)DCGANs:
题目:UNSUPERVISED REPRESENTATION LEARNING WITH DEEP CONVOLUTIONAL GENERATIVE ADVERSARIAL NETWORKS
文章链接:https://arxiv.org/pdf/1511.06434.pdf.
代码链接:https://github.com/Weifanwong/Implementation-of-classical-GANs-framework/tree/master/DCGAN.
目前效果最好的是CGANs和DCGANs,另外不得不说虽然仅仅刚接触了一点pytorch,感觉这个框架的确相比与tensorflow要方便得多,尤其在调试代码上。