Center for Data Innovation: “NVIDIA has released Flickr-Faces-HQ (FFHQ), a dataset of 70,000 high-resolution images of human faces. The dataset includes faces representing a wide range of ages and ethnicities, and the images also include humans wearing accessories such as eyeglasses, sunglasses, and hats. Researchers can use this dataset for multiple purposes, including training and testing generative adversarial networks. For example, NVIDIA used the dataset to develop StyleGan, an AI tool that generates realistic human faces.” Get the data.
- See also – A Style-Based Generator Architecture for Generative Adversarial Networks Tero Karras,Samuli Laine, Timo Aila (Submitted on 12 Dec 2018 (v1), last revised 6 Feb 2019 (this version, v2)) – “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.”