Progressive Growing of GANs for Improved Quality, Stability, and Variation
By Tero Karras, Timo Aila, Samuli Laine, Jaakko Lehtinen. 2017
8,667 citations recorded by the source corpus.
We describe a new training methodology for generative adversarial networks. The key idea is to grow both the generator and discriminator progressively: starting from a low resolution, we add new layers that model increasingly fine details as training progresses. This both speeds the training up and greatly stabilizes it, allowing us to produce images of unprecedented quality, e.g., CelebA images at 1024^2. We also propose a simple way to increase the variation in generated images, and achieve a record inception score of 8.80 in unsupervised CIFAR10. Additionally, we describe several implementation details that are important for discouraging unhealthy competition between the generator and discriminator. Finally, we suggest a new metric for evaluating GAN results, both in terms of image quality and variation. As an additional contribution, we construct a higher-quality version of the CelebA dataset.
Subfields: Computer Vision and Pattern Recognition, Generative Models
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Computer Vision and Pattern Recognition
- Luc Van Gool
- Dacheng Tao
- Ming-Hsuan Yang
- Trevor Darrell
- Xiaogang Wang
- Jitendra Malik
- Andrea Vedaldi
- Andrew Zisserman
In Generative Models
- Jun-Yan Zhu
- Stefano Ermon
- Eli Shechtman
- Richard Zhang
- Daniel Cohen-Or
- Arash Vahdat
- Jiaming Song
- Karsten Kreis
Citations in the corpus
- Denoising Diffusion Probabilistic Models (2020, background): “…adversarial networks (GANs), autoregressive models, flows, and variational autoencoders (VAEs) have synthesized striking image and audio samples [12, 25, 3, 55, 35, 23, 10, 30, 41, 54, 24, 31, 42], and there have been remarkable advances in energy-based modeling and score matching that have…”
- High-Resolution Image Synthesis with Latent Diffusion Models (2021, methodology): “On CelebA-HQ, we report a new state-of-the-art FID of 5.11, outperforming previous likelihood-based models as well as GANs.”
- Evaluating Lossy Compression Rates of Deep Generative Models (2020, background): “This may help explain the success of large GAN architectures (Brock et al., 2019; Karras et al., 2018a;b).”
- Improved Techniques for Training Score-Based Generative Models (2020, methodology): “As Table 1 shows (see the full table in Appendix C.1), our NCSNv2 achieves 37.3 on CelebA 64 2 which is comparable to ProgressiveGAN [13], whereas NCSN achieves 19.8.”
- Training Generative Adversarial Networks with Limited Data (2020, methodology): “We computed the FID for Progressive GAN [17] using the publicly available pre-trained network.”
- VQ-DRAW: A Sequential Discrete VAE (2020, background): “• The refinement network could be made "progressive" [19] [20], producing lower resolution outputs at earlier stages and higher resolution outputs at later stages.”
- CNN-Generated Images Are Surprisingly Easy to Spot… for Now (2019, methodology): “ProGAN [19] StyleGAN [20] BigGAN [7] CycleGAN [48] StarGAN [10] GauGAN [29] CRN [9] IMLE [23] SITD [8] Super-res.”
- Generative Modeling by Estimating Gradients of the Data Distribution (2019, neutral): “7 ProgressiveGAN [25] 8.”
- Flow++: Improving Flow-Based Generative Models with Variational Dequantization and Architecture Design (2019, background): “…begun to successfully model high dimensional raw observations from complex real-world datasets, from natural images and videos, to audio signals and natural language (Karras et al., 2017; Kalchbrenner et al., 2016b; van den Oord et al., 2016a; Kalchbrenner et al., 2016a; Vaswani et al., 2017).”
- A Style-Based Generator Architecture for Generative Adversarial Networks (2018, methodology): “The dataset includes vastly more variation than CelebA-HQ [28] in terms of age, ethnicity and image background, and also has much better coverage of accessories such as eyeglasses, sunglasses, hats, etc.”
- Large Scale GAN Training for High Fidelity Natural Image Synthesis (2018, methodology): “We find progressive growing (Karras et al., 2018) unnecessary even for our largest 512 × 512 models.”
- Generating High Fidelity Images with Subscale Pixel Networks and Multidimensional Upscaling (2018, methodology): “We produce CelebAHQ-256 samples (at full 8-bit resolution) that are of similar visual fidelity to those produced with methods such as GANs that lack however an intrinsic measure of generalization (Mescheder, 2018; Karras et al., 2017).”
- Glow: Generative Flow with Invertible 1x1 Convolutions (2018, methodology): “GANs (Goodfellow et al., 2014) are arguably best known for their ability to synthesize large and realistic images (Karras et al., 2017), in contrast with likelihood-based methods.”
- Autoregressive Quantile Networks for Generative Modeling (2018, background): “These have led to breakthroughs in state-of-the-art generation of natural images (Karras et al., 2017) and audio (van den Oord et al., 2016a), and even been used for unsupervised learning of disentangled representations (Hig-gins et al., 2017; Chen et al., 2016).”
- Generative Adversarial Networks (2021, neutral): “They used the same D architecture as in [29].”
- Towards Language-Free Training for Text-to-Image Generation (2021, background): “GANs [11] have inspired many advances in image synthesis [18, 20, 23, 28, 32].”
- Projected GANs Converge Faster (2021, background): “Providing multi-scale feedback with one or multiple discriminators has been helpful for both image synthesis [23, 24] and image-to-image translation [43, 64].”