All Projects → zalandoresearch → Psgan

zalandoresearch / Psgan

Licence: mit
Periodic Spatial Generative Adversarial Networks

Programming Languages

python
139335 projects - #7 most used programming language

Projects that are alternatives of or similar to Psgan

Generative adversarial networks 101
Keras implementations of Generative Adversarial Networks. GANs, DCGAN, CGAN, CCGAN, WGAN and LSGAN models with MNIST and CIFAR-10 datasets.
Stars: ✭ 138 (+27.78%)
Mutual labels:  jupyter-notebook, gan, dcgan, generative-adversarial-networks
Practical rl
A course in reinforcement learning in the wild
Stars: ✭ 4,741 (+4289.81%)
Mutual labels:  jupyter-notebook, lasagne, theano
Pytorch Pix2pix
Pytorch implementation of pix2pix for various datasets.
Stars: ✭ 74 (-31.48%)
Mutual labels:  gan, generative-model, generative-adversarial-networks
Gans In Action
Companion repository to GANs in Action: Deep learning with Generative Adversarial Networks
Stars: ✭ 748 (+592.59%)
Mutual labels:  jupyter-notebook, gan, dcgan
Tensorflow Generative Model Collections
Collection of generative models in Tensorflow
Stars: ✭ 3,785 (+3404.63%)
Mutual labels:  gan, generative-model, generative-adversarial-networks
Mimicry
[CVPR 2020 Workshop] A PyTorch GAN library that reproduces research results for popular GANs.
Stars: ✭ 458 (+324.07%)
Mutual labels:  gan, dcgan, generative-adversarial-networks
Deep Learning Python
Intro to Deep Learning, including recurrent, convolution, and feed forward neural networks.
Stars: ✭ 94 (-12.96%)
Mutual labels:  jupyter-notebook, lasagne, theano
Gan Tutorial
Simple Implementation of many GAN models with PyTorch.
Stars: ✭ 227 (+110.19%)
Mutual labels:  jupyter-notebook, gan, dcgan
Dcgan Tensorflow
A tensorflow implementation of "Deep Convolutional Generative Adversarial Networks"
Stars: ✭ 6,963 (+6347.22%)
Mutual labels:  gan, generative-model, dcgan
Discogan Pytorch
PyTorch implementation of "Learning to Discover Cross-Domain Relations with Generative Adversarial Networks"
Stars: ✭ 961 (+789.81%)
Mutual labels:  jupyter-notebook, gan, generative-model
Spectralnormalizationkeras
Spectral Normalization for Keras Dense and Convolution Layers
Stars: ✭ 100 (-7.41%)
Mutual labels:  jupyter-notebook, gan, generative-model
Generative models tutorial with demo
Generative Models Tutorial with Demo: Bayesian Classifier Sampling, Variational Auto Encoder (VAE), Generative Adversial Networks (GANs), Popular GANs Architectures, Auto-Regressive Models, Important Generative Model Papers, Courses, etc..
Stars: ✭ 276 (+155.56%)
Mutual labels:  jupyter-notebook, generative-model, dcgan
GAN-Project-2018
GAN in Tensorflow to be run via Linux command line
Stars: ✭ 21 (-80.56%)
Mutual labels:  gan, dcgan, generative-adversarial-networks
Ganotebooks
wgan, wgan2(improved, gp), infogan, and dcgan implementation in lasagne, keras, pytorch
Stars: ✭ 1,446 (+1238.89%)
Mutual labels:  jupyter-notebook, dcgan, lasagne
shoe-design-using-generative-adversarial-networks
No description or website provided.
Stars: ✭ 18 (-83.33%)
Mutual labels:  dcgan, generative-model, generative-adversarial-networks
Segan
Speech Enhancement Generative Adversarial Network in TensorFlow
Stars: ✭ 661 (+512.04%)
Mutual labels:  gan, generative-model, generative-adversarial-networks
Practical dl
DL course co-developed by YSDA, HSE and Skoltech
Stars: ✭ 1,006 (+831.48%)
Mutual labels:  jupyter-notebook, lasagne, theano
Catdcgan
A DCGAN that generate Cat pictures 🐱‍💻
Stars: ✭ 177 (+63.89%)
Mutual labels:  jupyter-notebook, gan, dcgan
Dragan
A stable algorithm for GAN training
Stars: ✭ 189 (+75%)
Mutual labels:  jupyter-notebook, gan, generative-model
Csc deeplearning
3-day dive into deep learning at csc
Stars: ✭ 22 (-79.63%)
Mutual labels:  jupyter-notebook, lasagne, theano

Periodic Spatial Generative Adversarial Networks

This code implements Periodic Spatial Generative Adversarial Networks (PSGANs) on top of Lasagne/Theano. The code was tested on top of Lasagne (version 0.2.dev1) and Theano (0.9.0dev2).

Update 13.12.2018

A Pytorch version of this texture synthesis method is now available at https://github.com/zalandoresearch/famos

Very Brief Model Description

PSGANs can generate sample textures of arbitrary size that look strikingly similar - but not exactly the same - compared to a single (or several) source image(s).

  • PSGANs can be thought of as a convolutional roll-out of Radford et al.'s deep convolutional generative adversarial networks for texture synthesis
  • the fully convolutional nature allows for real-time generation of high resolution images
  • the method can fuse multiple source images and is highly scalable w.r.t. to output texture size and allows for generation of tiled textures
  • can learn accurately periodic textures
  • can learn whole texture manifolds, allowing sampling and morphing between the training textures

Please take a look at our paper for detailed description of PSGANs https://arxiv.org/abs/1705.06566 .

Examples

NEW: better use of inference mode in the Ipython Notebook tutorial using pretrained statistics. In addition, a beatiful-looking pretrained model of DTD Scaly category is checked-in.

NEW: You can follow a step-by-step tutorial on inference and training PSGAN in the newly checked-in Ipython Notebook PSGAN tutorial for sampling and training.

You can generate samples from a stored model. E.g. you can use the checked-in model:

python demo_generation.py models/hex1_filters64_npx161_5gL_5dL_0Global_3Periodic_FalseAffine_30Local_epoch18.psgan

This model was trained on a single texture from the DTD Textures dataset, category Honeycomb https://www.robots.ox.ac.uk/~vgg/data/dtd/. Example image:

We have also provided an example of learning a whole texture manifold from the whole DTD Honeycomb category

python demo_generation.py models/honey_filters64_npx161_5gL_5dL_60Global_3Periodic_FalseAffine_30Local_epoch100.psgan

Training the Model

To train the model on new images, edit the config.py file and set the texture_dir variable to a folder containing the image(s) you want to learn from. You might also want to change other parameters of the network there (e.g. depth, number of filters per layer etc...). Then run

python psgan.py

to train the model. Snapshots of the model will be stored in the subfolder models/ and samples after each epoch will be stored in samples/

License

Please make sure to cite/acknowledge our paper, if you use any of the contained code in your own projects or publication.

The MIT License (MIT)

Copyright © 2017 Zalando SE, https://tech.zalando.com

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

Note that the project description data, including the texts, logos, images, and/or trademarks, for each open source project belongs to its rightful owner. If you wish to add or remove any projects, please contact us at [email protected].