OSInsert: Towards High-authenticity and High-fidelity Image Composition

TL;DR

OSInsert introduces a two-stage image composition method that combines high-authenticity foreground shape generation with high-fidelity detail preservation, achieving more realistic and detailed composite images.

Contribution

The paper proposes a novel two-stage strategy that jointly optimizes authenticity and fidelity in image composition, addressing limitations of existing methods.

Findings

Effective in generating realistic composite images

Verifies the approach on MureCOM dataset

Code and models publicly available

Abstract

Generative image composition aims to regenerate the given foreground object in the background image to produce a realistic composite image. Some high-authenticity methods can adjust foreground pose/view to be compatible with background, while some high-fidelity methods can preserve the foreground details accurately. However, existing methods can hardly achieve both goals at the same time. In this work, we propose a two-stage strategy to achieve both goals. In the first stage, we use high-authenticity method to generate reasonable foreground shape, serving as the condition of high-fidelity method in the second stage. The experiments on MureCOM dataset verify the effectiveness of our two-stage strategy. The code and model have been released at https://github.com/bcmi/OSInsert-Image-Composition.