BiGR: Harnessing Binary Latent Codes for Image Generation and Improved Visual Representation Capabilities
This work addresses the challenge of integrating generation and discrimination in vision models for researchers and practitioners, offering a novel framework with broad applications, though it builds on existing concepts like binary codes and masked modeling.
The paper tackles the problem of unifying generative and discriminative tasks in vision by introducing BiGR, a conditional image generation model using binary latent codes, which achieves superior generation quality (measured by FID-50k) and representation capabilities (measured by linear-probe accuracy) while enabling zero-shot generalization across tasks like inpainting and text-to-image generation.
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.