MAG-Edit: Localized Image Editing in Complex Scenarios via Mask-Based Attention-Adjusted Guidance

December 18, 2023 ยท Entered Twilight ยท ๐Ÿ› ACM Multimedia

๐Ÿ’ค TWILIGHT: Eternal Rest
Repo abandoned since publication

"No code URL or promise found in abstract"
"Derived repo from GitHub Pages (backfill)"

Evidence collected by the PWNC Scanner

Repo contents: README.md, assets, index.html, static, style.css

Authors Qi Mao, Lan Chen, Yuchao Gu, Zhen Fang, Mike Zheng Shou arXiv ID 2312.11396 Category cs.CV: Computer Vision Cross-listed cs.AI Citations 19 Venue ACM Multimedia Repository https://github.com/mag-edit/mag-edit.github.io โญ 1 Last Checked 9 days ago
Abstract
Recent diffusion-based image editing approaches have exhibited impressive editing capabilities in images with simple compositions. However, localized editing in complex scenarios has not been well-studied in the literature, despite its growing real-world demands. Existing mask-based inpainting methods fall short of retaining the underlying structure within the edit region. Meanwhile, mask-free attention-based methods often exhibit editing leakage and misalignment in more complex compositions. In this work, we develop MAG-Edit, a training-free, inference-stage optimization method, which enables localized image editing in complex scenarios. In particular, MAG-Edit optimizes the noise latent feature in diffusion models by maximizing two mask-based cross-attention constraints of the edit token, which in turn gradually enhances the local alignment with the desired prompt. Extensive quantitative and qualitative experiments demonstrate the effectiveness of our method in achieving both text alignment and structure preservation for localized editing within complex scenarios.
Community shame:
Not yet rated
Community Contributions

Found the code? Know the venue? Think something is wrong? Let us know!

๐Ÿ“œ Similar Papers

In the same crypt โ€” Computer Vision