🤖 AI Summary
Affective Image Manipulation (AIM) requires precise emotional elicitation while preserving the original image composition; however, existing methods rely solely on coarse-grained color or style adjustments, yielding superficial and imprecise emotional control. To address this, we propose a content-and-style co-editing framework. First, we introduce a psychology-informed content modification mechanism—the first of its kind in AIM. Second, we construct EmoEditSet, a large-scale paired dataset for emotion-guided editing. Third, we design EmoAdapter—a lightweight, plug-and-play module—and an instruction-aware loss function to enable fine-grained affective semantic modeling. Finally, we propose an emotion-attribution-driven data construction strategy to enhance semantic fidelity. Extensive experiments demonstrate significant improvements over state-of-the-art methods in both qualitative and quantitative evaluations. EmoAdapter is model-agnostic and seamlessly integrates with diverse diffusion models, consistently enhancing their affective expressiveness.
📝 Abstract
Affective Image Manipulation (AIM) seeks to modify user-provided images to evoke specific emotional responses. This task is inherently complex due to its twofold objective: significantly evoking the intended emotion, while preserving the original image composition. Existing AIM methods primarily adjust color and style, often failing to elicit precise and profound emotional shifts. Drawing on psychological insights, we introduce EmoEdit, which extends AIM by incorporating content modifications to enhance emotional impact. Specifically, we first construct EmoEditSet, a large-scale AIM dataset comprising 40,120 paired data through emotion attribution and data construction. To make existing generative models emotion-aware, we design the Emotion adapter and train it using EmoEditSet. We further propose an instruction loss to capture the semantic variations in data pairs. Our method is evaluated both qualitatively and quantitatively, demonstrating superior performance compared to existing state-of-the-art techniques. Additionally, we showcase the portability of our Emotion adapter to other diffusion-based models, enhancing their emotion knowledge with diverse semantics.