Progressive Checkerboards for Autoregressive Multiscale Image Generation
This work addresses efficiency in image generation for AI applications, presenting an incremental improvement over existing autoregressive systems.
The paper tackles the challenge of parallel sampling in autoregressive image generation by introducing a progressive checkerboard ordering that enables balanced conditioning across scales, achieving competitive performance on class-conditional ImageNet with fewer sampling steps compared to state-of-the-art methods.
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.