stylegan

## StyleGAN — Official TensorFlow Implementation **Picture:** *These people are not real – they were produced by our generator that allows control over different aspects of the image.* This repository contains the official TensorFlow implementation of the following paper: > **A Style-Based Generator Architecture for Generative Adversarial Networks**
> Tero Karras (NVIDIA), Samuli Laine (NVIDIA), Timo Aila (NVIDIA)
> https://arxiv.org/abs/1812.04948 > > **Abstract:** *We propose an alternative generator architecture for generative adversarial networks, borrowing from style transfer literature. The new architecture leads to an automatically learned, unsupervised separation of high-level attributes (e.g., pose and identity when trained on human faces) and stochastic variation in the generated images (e.g., freckles, hair), and it enables intuitive, scale-specific control of the synthesis. The new generator improves the state-of-the-art in terms of traditional distribution quality metrics, leads to demonstrably better interpolation properties, and also better disentangles the latent factors of variation. To quantify interpolation quality and disentanglement, we propose two new, automated methods that are applicable to any generator architecture. Finally, we introduce a new, highly varied and high-quality dataset of human faces.* For business inquiries, please visit our website and submit the form: [NVIDIA Research Licensing](https://www.nvidia.com/en-us/research/inquiries/) **★★★ NEW: [StyleGAN2-ADA-PyTorch](https://github.com/NVlabs/stylegan2-ada-pytorch) is now available; see the full list of versions [here](https://nvlabs.github.io/stylegan2/versions.html) ★★★** ## Resources Material related to our paper is available via the following links: - Paper: https://arxiv.org/abs/1812.04948 - Video: https://youtu.be/kSLJriaOumA - Code: https://github.com/NVlabs/stylegan - FFHQ: https://github.com/NVlabs/ffhq-dataset Additional material can be found on Google Drive: | Path | Description | :--- | :---------- | [StyleGAN](https://drive.google.com/open?id=1uka3a1noXHAydRPRbknqwKVGODvnmUBX) | Main folder. | ├  [stylegan-paper.pdf](https://drive.google.com/open?id=1v-HkF3Ehrpon7wVIx4r5DLcko_U_V6Lt) | High-quality version of the paper PDF. | ├  [stylegan-video.mp4](https://drive.google.com/open?id=1uzwkZHQX_9pYg1i0d1Nbe3D9xPO8-qBf) | High-quality version of the result video. | ├  [images](https://drive.google.com/open?id=1-l46akONUWF6LCpDoeq63H53rD7MeiTd) | Example images produced using our generator. | │  ├  [representative-images](https://drive.google.com/open?id=1ToY5P4Vvf5_c3TyUizQ8fckFFoFtBvD8) | High-quality images to be used in articles, blog posts, etc. | │  └  [100k-generated-images](https://drive.google.com/open?id=100DJ0QXyG89HZzB4w2Cbyf4xjNK54cQ1) | 100,000 generated images for different amounts of truncation. | │     ├  [ffhq-1024x1024](https://drive.google.com/open?id=14lm8VRN1pr4g_KVe6_LvyDX1PObst6d4) | Generated using Flickr-Faces-HQ dataset at 1024×1024. | │     ├  [bedrooms-256x256](https://drive.google.com/open?id=1Vxz9fksw4kgjiHrvHkX4Hze4dyThFW6t) | Generated using LSUN Bedroom dataset at 256×256. | │     ├  [cars-512x384](https://drive.google.com/open?id=1MFCvOMdLE2_mpeLPTiDw5dxc2CRuKkzS) | Generated using LSUN Car dataset at 512×384. | │     └  [cats-256x256](https://drive.google.com/open?id=1gq-Gj3GRFiyghTPKhp8uDMA9HV_0ZFWQ) | Generated using LSUN Cat dataset at 256×256. | ├  [videos](https://drive.google.com/open?id=1N8pOd_Bf8v89NGUaROdbD8-ayLPgyRRo) | Example videos produced using our generator. | │  └  [high-quality-video-clips](https://drive.google.com/open?id=1NFO7_vH0t98J13ckJYFd7kuaTkyeRJ86) | Individual segments of the result video as high-quality MP4. | ├  [ffhq-dataset](https://drive.google.com/open?id=1u2xu7bSrWxrbUxk-dT-UvEJq8IjdmNTP) | Raw data for the [Flickr-Faces-HQ dataset](https://github.com/NVlabs/ffhq-dataset). | └  [networks](https://drive.google.com/open?id=1MASQyN5m0voPcx7-9K0r5gObhvvPups7) | Pre-trained networks as pickled instances of [dnnlib.tflib.Network](./dnnlib/tflib/network.py). |    ├  [stylegan-ffhq-1024x1024.pkl](https://drive.google.com/uc?id=1MEGjdvVpUsu1jB4zrXZN7Y4kBBOzizDQ) | StyleGAN trained with Flickr-Faces-HQ dataset at 1024×1024. |    ├  [stylegan-celebahq-1024x1024.pkl](https://drive.google.com/uc?id=1MGqJl28pN4t7SAtSrPdSRJSQJqahkzUf) | StyleGAN trained with CelebA-HQ dataset at 1024×1024. |    ├  [stylegan-bedrooms-256x256.pkl](https://drive.google.com/uc?id=1MOSKeGF0FJcivpBI7s63V9YHloUTORiF) | StyleGAN trained with LSUN Bedroom dataset at 256×256. |    ├  [stylegan-cars-512x384.pkl](https://drive.google.com/uc?id=1MJ6iCfNtMIRicihwRorsM3b7mmtmK9c3) | StyleGAN trained with LSUN Car dataset at 512×384. |    ├  [stylegan-cats-256x256.pkl](https://drive.google.com/uc?id=1MQywl0FNt6lHu8E_EUqnRbviagS7fbiJ) | StyleGAN trained with LSUN Cat dataset at 256×256. |    └  [metrics](https://drive.google.com/open?id=1MvYdWCBuMfnoYGptRH-AgKLbPTsIQLhl) | Auxiliary networks for the quality and disentanglement metrics. |       ├  [inception_v3_features.pkl](https://drive.google.com/uc?id=1MzTY44rLToO5APn8TZmfR7_ENSe5aZUn) | Standard [Inception-v3](https://arxiv.org/abs/1512.00567) classifier that outputs a raw feature vector. |       ├  [vgg16_zhang_perceptual.pkl](https://drive.google.com/uc?id=1N2-m9qszOeVC9Tq77WxsLnuWwOedQiD2) | Standard [LPIPS](https://arxiv.org/abs/1801.03924) metric to estimate perceptual similarity. |       ├  [celebahq-classifier-00-male.pkl](https://drive.google.com/uc?id=1Q5-AI6TwWhCVM7Muu4tBM7rp5nG_gmCX) | Binary classifier trained to detect a single attribute of CelebA-HQ. |       └ ⋯ | Please see the file listing for remaining networks. ## Licenses All material, excluding the Flickr-Faces-HQ dataset, is made available under [Creative Commons BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/) license by NVIDIA Corporation. You can **use, redistribute, and adapt** the material for **non-commercial purposes**, as long as you give appropriate credit by **citing our paper** and **indicating any changes** that you've made. For license information regarding the FFHQ dataset, please refer to the [Flickr-Faces-HQ repository](https://github.com/NVlabs/ffhq-dataset). `inception_v3_features.pkl` and `inception_v3_softmax.pkl` are derived from the pre-trained [Inception-v3](https://arxiv.org/abs/1512.00567) network by Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jonathon Shlens, and Zbigniew Wojna. The network was originally shared under [Apache 2.0](https://github.com/tensorflow/models/blob/master/LICENSE) license on the [TensorFlow Models](https://github.com/tensorflow/models) repository. `vgg16.pkl` and `vgg16_zhang_perceptual.pkl` are derived from the pre-trained [VGG-16](https://arxiv.org/abs/1409.1556) network by Karen Simonyan and Andrew Zisserman. The network was originally shared under [Creative Commons BY 4.0](https://creativecommons.org/licenses/by/4.0/) license on the [Very Deep Convolutional Networks for Large-Scale Visual Recognition](http://www.robots.ox.ac.uk/~vgg/research/very_deep/) project page. `vgg16_zhang_perceptual.pkl` is further derived from the pre-trained [LPIPS](https://arxiv.org/abs/1801.03924) weights by Richard Zhang, Phillip Isola, Alexei A. Efros, Eli Shechtman, and Oliver Wang. The weights were originally shared under [BSD 2-Clause "Simplified" License](https://github.com/richzhang/PerceptualSimilarity/blob/master/LICENSE) on the [PerceptualSimilarity](https://github.com/richzhang/PerceptualSimilarity) repository. ## System requirements * Both Linux and Windows are supported, but we strongly recommend Linux for performance and compatibility reasons. * 64-bit Python 3.6 installation. We recommend Anaconda3 with numpy 1.14.3 or newer. * TensorFlow 1.10.0 or newer with GPU support. * One or more high-end NVIDIA GPUs with at least 11GB of DRAM. We recommend NVIDIA DGX-1 with 8 Tesla V100 GPUs. * NVIDIA driver 391.35 or newer, CUDA toolkit 9.0 or newer, cuDNN 7.3.1 or newer. ## Using pre-trained networks A minimal example of using a pre-trained StyleGAN generator is given in [pretrained_example.py](./pretrained_example.py). When executed, the script downloads a pre-trained StyleGAN generator from Google Drive and uses it to generate an image: ``` … ``` A more advanced example is given in [generate_figures.py](./generate_figures.py). The script reproduces the figures from our paper in order to illustrate style mixing, noise inputs, and truncation: ``` … ``` The pre-trained networks are stored as standard pickle files on Google Drive: ``` # Load pre-trained network. url = 'https://drive.google.com/uc?id=1MEGjdvVpUsu1jB4zrXZN7Y4kBBOzizDQ' # karras2019stylegan-ffhq-1024x1024.pkl with dnnlib.util.open_url(url, cache_dir=config.cache_dir) as f: _G, _D, Gs = pickle.load(f) # _G = Instantaneous snapshot of the generator. Mainly useful for resuming a previous training run. # _D = Instantaneous snapshot of the discriminator. Mainly useful for resuming a previous training run. # Gs = Long-term average of the generator. Yields higher-quality results than the instantaneous snapshot. ``` The above code downloads the file and unpickles it to yield 3 instances of [dnnlib.tflib.Network](./dnnlib/tflib/network.py). To generate images, you will typically want to use `Gs` – the other two networks are provided for completeness. In order for `pickle.load()` to work, you will need to have the `dnnlib` source directory in your PYTHONPATH and a `tf.Session` set as default. The session can initialized by calling `dnnlib.tflib.init_tf()`. There are three ways to use the pre-trained generator: 1. Use `Gs.run()` for immediate-mode operation where the inputs and outputs are numpy arrays: ``` # Pick latent vector. rnd = np.random.RandomState(5) latents = rnd.randn(1, Gs.input_shape[1]) # Generate image. fmt = dict(func=tflib.convert_images_to_uint8, nchw_to_nhwc=True) images = Gs.run(latents, None, truncation_psi=0.7, randomize_noise=True, output_transform=fmt) ``` The first argument is a batch of latent vectors of shape `[num, 512]`. The second argument is reserved for class labels (not used by StyleGAN). The remaining keyword arguments are optional and can be used to further modify the operation (see below). The output is a batch of images, whose format is dictated by the `output_transform` argument. 2. Use `Gs.get_output_for()` to incorporate the generator as a part of a larger TensorFlow expression: ``` latents = tf.random_normal([self.minibatch_per_gpu] + Gs_clone.input_shape[1:]) images = Gs_clone.get_output_for(latents, None, is_validation=True, randomize_noise=True) images = tflib.convert_images_to_uint8(images) result_expr.append(inception_clone.get_output_for(images)) ``` The above code is from [metrics/frechet_inception_distance.py](./metrics/frechet_inception_distance.py). It generates a batch of random images and feeds them directly to the [Inception-v3](https://arxiv.org/abs/1512.00567) network without having to convert the data to numpy arrays in between. 3. Look up `Gs.components.mapping` and `Gs.components.synthesis` to access individual sub-networks of the generator. Similar to `Gs`, the sub-networks are represented as independent instances of [dnnlib.tflib.Network](./dnnlib/tflib/