PixWizard: Versatile Image-to-Image Visual Assistant with Open-Language Instructions

TL;DR

PixWizard is a versatile image-to-image assistant that unifies multiple vision tasks into a single framework, enabling flexible, instruction-driven image generation and manipulation across diverse resolutions and tasks.

Contribution

The paper introduces PixWizard, a unified model for diverse vision tasks guided by natural language instructions, with a large instruction-tuning dataset and a flexible resolution mechanism.

Findings

Demonstrates strong generalization to unseen tasks and instructions.

Achieves high-quality image generation and editing across various resolutions.

Shows effective fusion of image structure and semantics.

Abstract

This paper presents a versatile image-to-image visual assistant, PixWizard, designed for image generation, manipulation, and translation based on free-from language instructions. To this end, we tackle a variety of vision tasks into a unified image-text-to-image generation framework and curate an Omni Pixel-to-Pixel Instruction-Tuning Dataset. By constructing detailed instruction templates in natural language, we comprehensively include a large set of diverse vision tasks such as text-to-image generation, image restoration, image grounding, dense image prediction, image editing, controllable generation, inpainting/outpainting, and more. Furthermore, we adopt Diffusion Transformers (DiT) as our foundation model and extend its capabilities with a flexible any resolution mechanism, enabling the model to dynamically process images based on the aspect ratio of the input, closely aligning with human perceptual processes. The model also incorporates structure-aware and semantic-aware guidance to facilitate effective fusion of information from the input image. Our experiments demonstrate that PixWizard not only shows impressive generative and understanding abilities for images with diverse resolutions but also exhibits promising generalization capabilities with unseen tasks and human instructions. The code and related resources are available at https://github.com/AFeng-x/PixWizard