Progressive Checkerboards for Autoregressive Multiscale Image Generation

TL;DR

This paper introduces a progressive checkerboard approach for multiscale autoregressive image generation, enabling efficient parallel sampling while maintaining dependencies, and achieves competitive results with fewer steps.

Contribution

It proposes a flexible, fixed ordering based on progressive checkerboards that balances parallel sampling and dependency modeling in multiscale image generation.

Findings

Effective conditioning within and between scales.

Similar results across various scale-up factors with constant serial steps.

Competitive performance on class-conditional ImageNet.

Abstract

A key challenge in autoregressive image generation is to efficiently sample independent locations in parallel, while still modeling mutual dependencies with serial conditioning. Some recent works have addressed this by conditioning between scales in a multiscale pyramid. Others have looked at parallelizing samples in a single image using regular partitions or randomized orders. In this work we examine a flexible, fixed ordering based on progressive checkerboards for multiscale autoregressive image generation. Our ordering draws samples in parallel from evenly spaced regions at each scale, maintaining full balance in all levels of a quadtree subdivision at each step. This enables effective conditioning both between and within scales. Intriguingly, we find evidence that in our balanced setting, a wide range of scale-up factors lead to similar results, so long as the total number of serial steps is constant. On class-conditional ImageNet, our method achieves competitive performance compared to recent state-of-the-art autoregressive systems with like model capacity, using fewer sampling steps.