A Style-Based Generator Architecture for Generative Adversarial Networks
By Tero Karras, Samuli Laine, Timo Aila. 2018
13,265 citations recorded by the source corpus.
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.
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
- Eli Shechtman
- Richard Zhang
- Daniel Cohen-Or
- Stefano Ermon
- Arash Vahdat
- Alexei A. Efros
- Jiaming Song
Citations in the corpus
- High-Resolution Image Synthesis with Latent Diffusion Models (2021, methodology): “Samples from LDMs trained on CelebAHQ [38], FFHQ [40], LSUN-Churches [98], LSUN-Bedrooms [98] and class-conditional ImageNet [12], each with a resolution of 256× 256.”
- FairFace: Face Attribute Dataset for Balanced Race, Gender, and Age (2019, background): “To date, numerous large scale face image datasets [22, 31, 13, 69, 37, 24, 41, 68, 15, 27, 46, 8, 39] have been proposed and fostered research and development for automated face detection [35, 21], alignment [66, 44], recognition [57, 49], generation [67, 5, 26, 58], modification [3, 32, 19], and at”
- Making Convolutional Networks Shift-Invariant Again (2019, methodology): “Our classification experiments indicate that anti-aliasing is a natural choice for the discriminator, and is used in the recent StyleGAN method (Karras et al., 2019).”
- 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): “StyleGAN∗ [28] 50.”
- Training Generative Adversarial Networks with Limited Data (2020, background): “The increasingly impressive results of generative adversarial networks (GAN) [14, 32, 31, 5, 19, 20, 21] are fueled by the seemingly unlimited supply of images available online.”
- 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.”
- Analyzing and Improving the Image Quality of StyleGAN (2019, methodology): “This corresponds to configuration E in Table 1, and it signifi-cantly improves FID and PPL.”
- Generative Modeling by Estimating Gradients of the Data Distribution (2019, background): “To list a few, they have been used to generate high-fidelity images [26, 6], synthesize realistic speech and music fragments [58], improve the performance of semi-supervised learning [28, 10], detect adversarial examples and other anomalous data [54], imitation learning [22], and explore promising s”
- Generating Diverse High-Fidelity Images with VQ-VAE-2 (2019, methodology): “To further assess the effectiveness of our multi-scale approach for capturing extremely long range dependencies in the data, we train a three level hierarchical model over the FFHQ dataset [15] at 1024 × 1024 resolution.”
- GLIDE: Towards Photorealistic Image Generation and Editing with Text-Guided Diffusion Models (2021, background): “On the other hand, unconditional image models can synthesize photorealistic images (Brock et al., 2018; Karras et al., 2019a;b; Razavi et al., 2019), sometimes with enough fidelity that humans can’t distinguish them from real images (Zhou et al., 2019).”
- Generative Adversarial Networks (2021, neutral): “BigGAN is capable of generating high-quality images with up to 512x512 pixels.”
- Towards Language-Free Training for Text-to-Image Generation (2021, methodology): “For face domain, we pre-trained a model on FFHQ dataset [19] which contains 70,000 images.”
- Projected GANs Converge Faster (2021, methodology): “On the large datasets, the baselines only outperform Projected GAN in precision on FFHQ, Cityscapes, and LSUN Church (Table 4).”
- Vector-quantized Image Modeling with Improved VQGAN (2021, methodology): “After the sweep, we apply the same set of hyper-parameters of training losses to all datasets including CelebA-HQ, FFHQ, and ImageNet.”
- Resolution-robust Large Mask Inpainting with Fourier Convolutions (2021, methodology): “Following previous works [31, 22], we use an r1-gradient penalized patch discriminator in our system.”
- ImageBART: Bidirectional Context with Multinomial Diffusion for Autoregressive Image Synthesis (2021, methodology): “We train models at resolution 256 × 256 for unconditional generation on FFHQ [30], LSUN -Cats, -Churches and -Bedrooms [76] and on class-conditional synthesis on ImageNet (cIN) [11].”
- D2C: Diffusion-Decoding Models for Few-Shot Conditional Generation (2021, background): “We examine the conditional and unconditional generation qualities of D2C over CIFAR-10 [50], CIFAR-100 [50], fMoW [19], CelebA-64 [55], CelebA-HQ-256 [45], and FFHQ-256 [46].”
- High-Resolution Complex Scene Synthesis with Transformers (2021, methodology): “The generator is conditioned on the scene layout and mask predictors via a layout-aware norm, inspired by StyleGAN [15].”