3D-Fixup: Advancing Photo Editing with 3D Priors

TL;DR

3D-Fixup introduces a framework that leverages learned 3D priors and diffusion models to enable realistic, identity-preserving 3D-aware edits of 2D images, including object translation and rotation.

Contribution

It presents a novel training-based approach that uses video data and 3D guidance from an Image-to-3D model to improve 3D-aware image editing capabilities.

Findings

Supports complex 3D transformations like translation and rotation

Achieves high-quality, identity-coherent edits

Effectively integrates 3D priors with diffusion models

Abstract

Despite significant advances in modeling image priors via diffusion models, 3D-aware image editing remains challenging, in part because the object is only specified via a single image. To tackle this challenge, we propose 3D-Fixup, a new framework for editing 2D images guided by learned 3D priors. The framework supports difficult editing situations such as object translation and 3D rotation. To achieve this, we leverage a training-based approach that harnesses the generative power of diffusion models. As video data naturally encodes real-world physical dynamics, we turn to video data for generating training data pairs, i.e., a source and a target frame. Rather than relying solely on a single trained model to infer transformations between source and target frames, we incorporate 3D guidance from an Image-to-3D model, which bridges this challenging task by explicitly projecting 2D information into 3D space. We design a data generation pipeline to ensure high-quality 3D guidance throughout training. Results show that by integrating these 3D priors, 3D-Fixup effectively supports complex, identity coherent 3D-aware edits, achieving high-quality results and advancing the application of diffusion models in realistic image manipulation. The code is provided at https://3dfixup.github.io/