Generating images with recurrent adversarial networks

TL;DR

This document presents a comprehensive style guide for CVPR submissions, detailing language requirements, submission policies, page-length constraints, and detailed formatting rules. It emphasizes a standardized two-column layout, precise margins, font usage, and equation numbering, along with strict blind-review practices and citation standards. The guide also covers graphical content, color usage, and the finalCopyright process to ensure consistent, fair, and reproducible conference submissions. Overall, it serves to streamline the review process and ensure high-quality, uniformly formatted proceedings materials.

Abstract

Gatys et al. (2015) showed that optimizing pixels to match features in a convolutional network with respect reference image features is a way to render images of high visual quality. We show that unrolling this gradient-based optimization yields a recurrent computation that creates images by incrementally adding onto a visual "canvas". We propose a recurrent generative model inspired by this view, and show that it can be trained using adversarial training to generate very good image samples. We also propose a way to quantitatively compare adversarial networks by having the generators and discriminators of these networks compete against each other.

Figures (2)

• Figure 1: Example of caption. It is set in Roman so that mathematics (always set in Roman: $B \sin A = A \sin B$) may be included without an ugly clash.
• Figure 2: Example of a short caption, which should be centered.