OSInsert: Towards High-authenticity and High-fidelity Image Composition
This work addresses the challenge of creating realistic composite images for applications in computer vision and graphics, though it appears incremental as it combines existing high-authenticity and high-fidelity methods.
The paper tackles the problem of generative image composition by proposing a two-stage strategy to achieve both high authenticity and high fidelity, which existing methods struggle to do simultaneously, and experiments on the MureCOM dataset verify its effectiveness.
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.