EditWorld: Simulating World Dynamics for Instruction-Following Image Editing

TL;DR

EditWorld introduces a novel instruction-based image editing task focused on simulating world dynamics, utilizing a curated dataset and training strategies to enhance realistic, dynamic scene editing beyond simple modifications.

Contribution

The paper presents a new world-instructed image editing task, a curated dataset with world instructions, and a training method that improves instruction-following for dynamic scene editing.

Findings

Outperforms existing editing methods in the new task

Demonstrates effective simulation of world dynamics in image editing

Provides a dataset and code for further research

Abstract

Diffusion models have significantly improved the performance of image editing. Existing methods realize various approaches to achieve high-quality image editing, including but not limited to text control, dragging operation, and mask-and-inpainting. Among these, instruction-based editing stands out for its convenience and effectiveness in following human instructions across diverse scenarios. However, it still focuses on simple editing operations like adding, replacing, or deleting, and falls short of understanding aspects of world dynamics that convey the realistic dynamic nature in the physical world. Therefore, this work, EditWorld, introduces a new editing task, namely world-instructed image editing, which defines and categorizes the instructions grounded by various world scenarios. We curate a new image editing dataset with world instructions using a set of large pretrained models (e.g., GPT-3.5, Video-LLava and SDXL). To enable sufficient simulation of world dynamics for image editing, our EditWorld trains model in the curated dataset, and improves instruction-following ability with designed post-edit strategy. Extensive experiments demonstrate our method significantly outperforms existing editing methods in this new task. Our dataset and code will be available at https://github.com/YangLing0818/EditWorld