Addressing Text Embedding Leakage in Diffusion-based Image Editing

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Mun, Sunung, Nam, Jinhwan, Cho, Sunghyun, Ok, Jungseul
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915458443116544
author Mun, Sunung
Nam, Jinhwan
Cho, Sunghyun
Ok, Jungseul
author_facet Mun, Sunung
Nam, Jinhwan
Cho, Sunghyun
Ok, Jungseul
contents Text-based image editing, powered by generative diffusion models, lets users modify images through natural-language prompts and has dramatically simplified traditional workflows. Despite these advances, current methods still suffer from a critical problem: attribute leakage, where edits meant for specific objects unintentionally affect unrelated regions or other target objects. Our analysis reveals the root cause as the semantic entanglement inherent in End-of-Sequence (EOS) embeddings generated by autoregressive text encoders, which indiscriminately aggregate attributes across prompts. To address this issue, we introduce Attribute-Leakage-free Editing (ALE), a framework that tackles attribute leakage at its source. ALE combines Object-Restricted Embeddings (ORE) to disentangle text embeddings, Region-Guided Blending for Cross-Attention Masking (RGB-CAM) for spatially precise attention, and Background Blending (BB) to preserve non-edited content. To quantitatively evaluate attribute leakage across various editing methods, we propose the Attribute-Leakage Evaluation Benchmark (ALE-Bench), featuring comprehensive editing scenarios and new metrics. Extensive experiments show that ALE reduces attribute leakage by large margins, thereby enabling accurate, multi-object, text-driven image editing while faithfully preserving non-target content.
format Preprint
id arxiv_https___arxiv_org_abs_2412_04715
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Addressing Text Embedding Leakage in Diffusion-based Image Editing
Mun, Sunung
Nam, Jinhwan
Cho, Sunghyun
Ok, Jungseul
Computer Vision and Pattern Recognition
Text-based image editing, powered by generative diffusion models, lets users modify images through natural-language prompts and has dramatically simplified traditional workflows. Despite these advances, current methods still suffer from a critical problem: attribute leakage, where edits meant for specific objects unintentionally affect unrelated regions or other target objects. Our analysis reveals the root cause as the semantic entanglement inherent in End-of-Sequence (EOS) embeddings generated by autoregressive text encoders, which indiscriminately aggregate attributes across prompts. To address this issue, we introduce Attribute-Leakage-free Editing (ALE), a framework that tackles attribute leakage at its source. ALE combines Object-Restricted Embeddings (ORE) to disentangle text embeddings, Region-Guided Blending for Cross-Attention Masking (RGB-CAM) for spatially precise attention, and Background Blending (BB) to preserve non-edited content. To quantitatively evaluate attribute leakage across various editing methods, we propose the Attribute-Leakage Evaluation Benchmark (ALE-Bench), featuring comprehensive editing scenarios and new metrics. Extensive experiments show that ALE reduces attribute leakage by large margins, thereby enabling accurate, multi-object, text-driven image editing while faithfully preserving non-target content.
title Addressing Text Embedding Leakage in Diffusion-based Image Editing
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2412.04715