BiGR: Harnessing Binary Latent Codes for Image Generation and Improved Visual Representation Capabilities

TL;DR

BiGR introduces a unified binary latent code-based model for high-quality image generation and versatile visual representations, achieving superior performance and zero-shot generalization across multiple vision tasks.

Contribution

It is the first to unify generation and discrimination within a single framework using binary latent codes and introduces novel mechanisms like entropy-ordered sampling.

Findings

BiGR achieves state-of-the-art FID-50k scores.

It demonstrates high linear-probe accuracy for representations.

Enables zero-shot tasks like inpainting, outpainting, and editing.

Abstract

We introduce BiGR, a novel conditional image generation model using compact binary latent codes for generative training, focusing on enhancing both generation and representation capabilities. BiGR is the first conditional generative model that unifies generation and discrimination within the same framework. BiGR features a binary tokenizer, a masked modeling mechanism, and a binary transcoder for binary code prediction. Additionally, we introduce a novel entropy-ordered sampling method to enable efficient image generation. Extensive experiments validate BiGR's superior performance in generation quality, as measured by FID-50k, and representation capabilities, as evidenced by linear-probe accuracy. Moreover, BiGR showcases zero-shot generalization across various vision tasks, enabling applications such as image inpainting, outpainting, editing, interpolation, and enrichment, without the need for structural modifications. Our findings suggest that BiGR unifies generative and discriminative tasks effectively, paving the way for further advancements in the field. We further enable BiGR to perform text-to-image generation, showcasing its potential for broader applications.