Deep Generative Image Models using a Laplacian Pyramid of Adversarial Networks

TLDR

The paper introduces a generative parametric model that produces high‑quality natural images. The model uses a cascade of convolutional networks in a Laplacian pyramid, training each level with a GAN to generate images from coarse to fine. Human evaluation shows the model’s CIFAR‑10 samples are mistaken for real images 40% of the time, compared to 10% for a baseline GAN, indicating substantially higher quality.

Abstract

In this paper we introduce a generative parametric model capable of producing high quality samples of natural images. Our approach uses a cascade of convolutional networks within a Laplacian pyramid framework to generate images in a coarse-to-fine fashion. At each level of the pyramid, a separate generative convnet model is trained using the Generative Adversarial Nets (GAN) approach (Goodfellow et al.). Samples drawn from our model are of significantly higher quality than alternate approaches. In a quantitative assessment by human evaluators, our CIFAR10 samples were mistaken for real images around 40% of the time, compared to 10% for samples drawn from a GAN baseline model. We also show samples from models trained on the higher resolution images of the LSUN scene dataset.