Sense it with your eyes: Sensation Generation and Understanding for Advertisements

Aysan Aghazadeh, Adriana Kovashka

Keywords: Image Generation, Image Understanding, Sensory Images, Advertisement Images, Persuasion

Abstract: Effective advertisements might persuade the audience by evoking human sensations, yet current Text-to-Image (T2I) models struggle to generate persuasive visuals that convey implicit sensory experiences. We introduce the SensoryAd Generation task: given an advertisement message and a specific sensation related to the advertisement message, the goal is to generate advertisement images that both convey the message and evoke the sensation. To support this task, we build the SensoryAd dataset, consisting of human-designed and generated advertisements annotated with sensation categories, visual elements evoking the sensation, and human ratings. We further propose a novel evaluation method using contrastive and consistency losses across hierarchical sensation levels.

Primary Area: datasets and benchmarks

Submission Number: 3488