A Style-Based Generator Architecture for Generative Adversarial Networks
Top Cited Papers
- 1 June 2019
- conference paper
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 4396-4405
- https://doi.org/10.1109/cvpr.2019.00453
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.Keywords
This publication has 11 references indexed in Scilit:
- Feature-wise transformationsDistill, 2018
- MIXGAN: Learning Concepts from Different Domains for Mixture GenerationPublished by International Joint Conferences on Artificial Intelligence ,2018
- The Unreasonable Effectiveness of Deep Features as a Perceptual MetricPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2018
- Arbitrary Style Transfer in Real-Time with Adaptive Instance NormalizationPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2017
- Demystifying Neural Style TransferPublished by International Joint Conferences on Artificial Intelligence ,2017
- Image Style Transfer Using Convolutional Neural NetworksPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2016
- One millisecond face alignment with an ensemble of regression treesPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2014
- Learning Factorial Codes by Predictability MinimizationNeural Computation, 1992
- Improving generalization performance using double backpropagationIEEE Transactions on Neural Networks, 1992
- Animating rotation with quaternion curvesPublished by Association for Computing Machinery (ACM) ,1985