Delta-Adapter: Scalable Exemplar-Based Image Editing with Single-Pair Supervision

TL;DR

Delta-Adapter introduces a scalable image editing approach that learns from single image pairs by extracting semantic deltas with a pre-trained encoder, enabling effective transfer of editing semantics without requiring multiple exemplar pairs.

Contribution

It proposes a novel single-pair supervision method using semantic deltas and a Perceiver-based adapter, improving generalization and editing accuracy over existing multi-pair supervised methods.

Findings

Outperforms four strong baselines on seen editing tasks.

Generalizes more effectively to unseen editing tasks.

Leverages large-scale datasets without needing multiple exemplar pairs.

Abstract

Exemplar-based image editing applies a transformation defined by a source-target image pair to a new query image. Existing methods rely on a pair-of-pairs supervision paradigm, requiring two image pairs sharing the same edit semantics to learn the target transformation. This constraint makes training data difficult to curate at scale and limits generalization across diverse edit types. We propose Delta-Adapter, a method that learns transferable editing semantics under single-pair supervision, requiring no textual guidance. Rather than directly exposing the exemplar pair to the model, we leverage a pre-trained vision encoder to extract a semantic delta that encodes the visual transformation between the two images. This semantic delta is injected into a pre-trained image editing model via a Perceiver-based adapter. Since the target image is never directly visible to the model, it can serve as the prediction target, enabling single-pair supervision without requiring additional exemplar pairs. This formulation allows us to leverage existing large-scale editing datasets for training. To further promote faithful transformation transfer, we introduce a semantic delta consistency loss that aligns the semantic change of the generated output with the ground-truth semantic delta extracted from the exemplar pair. Extensive experiments demonstrate that Delta-Adapter consistently improves both editing accuracy and content consistency over four strong baselines on seen editing tasks, while also generalizing more effectively to unseen editing tasks. Code will be available at https://delta-adapter.github.io.