PairEdit: Learning Semantic Variations for Exemplar-based Image Editing

2025-06-09

AI Summary

Existing image editing methods heavily rely on textual prompts to describe editing semantics, making precise characterization of complex visual transformations challenging. This paper introduces the first purely image-driven few-shot (even single-pair) semantic editing framework, requiring no textual guidance. Our method leverages diffusion models to achieve high-fidelity editing. Key contributions include: (1) a target noise prediction mechanism that explicitly models the semantic transformation direction between image pairs; (2) a content-preserving adaptive noise scheduling strategy to enhance structural consistency; and (3) a dual-LoRA decoupled training scheme that separately learns editing style and content representations. Extensive qualitative and quantitative evaluations demonstrate state-of-the-art performance, with significantly improved content consistency over existing baselines.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Search and Optimization: Learning to Search
• Natural Language Processing: Sentence-level Semantics, Textual Inference, etc.

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Recent advancements in text-guided image editing have achieved notable success by leveraging natural language prompts for fine-grained semantic control. However, certain editing semantics are challenging to specify precisely using textual descriptions alone. A practical alternative involves learning editing semantics from paired source-target examples. Existing exemplar-based editing methods still rely on text prompts describing the change within paired examples or learning implicit text-based editing instructions. In this paper, we introduce PairEdit, a novel visual editing method designed to effectively learn complex editing semantics from a limited number of image pairs or even a single image pair, without using any textual guidance. We propose a target noise prediction that explicitly models semantic variations within paired images through a guidance direction term. Moreover, we introduce a content-preserving noise schedule to facilitate more effective semantic learning. We also propose optimizing distinct LoRAs to disentangle the learning of semantic variations from content. Extensive qualitative and quantitative evaluations demonstrate that PairEdit successfully learns intricate semantics while significantly improving content consistency compared to baseline methods. Code will be available at https://github.com/xudonmao/PairEdit.

Problem

• Learning editing semantics from image pairs without text
• Modeling semantic variations in paired images explicitly
• Disentangling semantic learning from content for better consistency

Innovation

• Learns editing semantics from image pairs
• Uses target noise prediction for variations
• Optimizes LoRAs to disentangle semantic learning