Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Gong, Yan, Song, Yiren, Li, Yicheng, Li, Chenglin, Zhang, Yin
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2506.02528
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866912411110342656
author Gong, Yan
Song, Yiren
Li, Yicheng
Li, Chenglin
Zhang, Yin
author_facet Gong, Yan
Song, Yiren
Li, Yicheng
Li, Chenglin
Zhang, Yin
contents Inspired by the in-context learning mechanism of large language models (LLMs), a new paradigm of generalizable visual prompt-based image editing is emerging. Existing single-reference methods typically focus on style or appearance adjustments and struggle with non-rigid transformations. To address these limitations, we propose leveraging source-target image pairs to extract and transfer content-aware editing intent to novel query images. To this end, we introduce RelationAdapter, a lightweight module that enables Diffusion Transformer (DiT) based models to effectively capture and apply visual transformations from minimal examples. We also introduce Relation252K, a comprehensive dataset comprising 218 diverse editing tasks, to evaluate model generalization and adaptability in visual prompt-driven scenarios. Experiments on Relation252K show that RelationAdapter significantly improves the model's ability to understand and transfer editing intent, leading to notable gains in generation quality and overall editing performance.
format Preprint
id arxiv_https___arxiv_org_abs_2506_02528
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle RelationAdapter: Learning and Transferring Visual Relation with Diffusion Transformers
Gong, Yan
Song, Yiren
Li, Yicheng
Li, Chenglin
Zhang, Yin
Computer Vision and Pattern Recognition
Inspired by the in-context learning mechanism of large language models (LLMs), a new paradigm of generalizable visual prompt-based image editing is emerging. Existing single-reference methods typically focus on style or appearance adjustments and struggle with non-rigid transformations. To address these limitations, we propose leveraging source-target image pairs to extract and transfer content-aware editing intent to novel query images. To this end, we introduce RelationAdapter, a lightweight module that enables Diffusion Transformer (DiT) based models to effectively capture and apply visual transformations from minimal examples. We also introduce Relation252K, a comprehensive dataset comprising 218 diverse editing tasks, to evaluate model generalization and adaptability in visual prompt-driven scenarios. Experiments on Relation252K show that RelationAdapter significantly improves the model's ability to understand and transfer editing intent, leading to notable gains in generation quality and overall editing performance.
title RelationAdapter: Learning and Transferring Visual Relation with Diffusion Transformers
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2506.02528