The same ideas used for forging MNIST images can be applied to other image domains. In this recipe, you will learn how to use the package located at https://github.com/carpedm20/DCGAN-tensorflow to train a DCGAN model on different datasets. The work is based on the paper Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks, Alec Radford, Luke Metz, Soumith Chintal, 2015. Quoting the abstract:
In recent years, supervised learning with convolutional networks (CNNs) has seen huge adoption in computer vision applications. Comparatively, unsupervised learning with CNNs has received less attention. In this work we hope to help bridge the gap between the success of CNNs for supervised learning and unsupervised learning. We introduce a class of CNNs called deep convolutional generative adversarial networks (DCGANs), that have certain architectural constraints, and demonstrate that they are a strong candidate for unsupervised learning. Training on various image datasets, we show convincing evidence that our deep convolutional adversarial pair learns a hierarchy of representations from object parts to scenes in both the generator and discriminator. Additionally, we use the learned features for novel tasks - demonstrating their applicability as general image representations.
Note that the generator has the architecture represented in the following diagram:
Note that in the package there are changes with respect to the original paper to avoid the fast convergence of D (discriminator) network, G (generator) network is updated twice for each D network update.