ImageBART: Bidirectional Context with Multinomial Diffusion for Autoregressive Image Synthesis

TL;DR

ImageBART introduces a hierarchical autoregressive model that combines multinomial diffusion with a coarse-to-fine approach, enabling more global context understanding and improved image editing capabilities.

Contribution

It presents a novel combination of autoregressive modeling with multinomial diffusion to incorporate global context in image synthesis and editing.

Findings

Enhanced image modification capabilities over traditional autoregressive models

High-fidelity image generation in a compressed latent space

Effective local image editing and inpainting without mask-specific training

Abstract

Autoregressive models and their sequential factorization of the data likelihood have recently demonstrated great potential for image representation and synthesis. Nevertheless, they incorporate image context in a linear 1D order by attending only to previously synthesized image patches above or to the left. Not only is this unidirectional, sequential bias of attention unnatural for images as it disregards large parts of a scene until synthesis is almost complete. It also processes the entire image on a single scale, thus ignoring more global contextual information up to the gist of the entire scene. As a remedy we incorporate a coarse-to-fine hierarchy of context by combining the autoregressive formulation with a multinomial diffusion process: Whereas a multistage diffusion process successively removes information to coarsen an image, we train a (short) Markov chain to invert this process. In each stage, the resulting autoregressive ImageBART model progressively incorporates context from previous stages in a coarse-to-fine manner. Experiments show greatly improved image modification capabilities over autoregressive models while also providing high-fidelity image generation, both of which are enabled through efficient training in a compressed latent space. Specifically, our approach can take unrestricted, user-provided masks into account to perform local image editing. Thus, in contrast to pure autoregressive models, it can solve free-form image inpainting and, in the case of conditional models, local, text-guided image modification without requiring mask-specific training.