AdBooster: Personalized Ad Creative Generation using Stable Diffusion Outpainting

Veronika Shilova,Ludovic Dos Santos,Flavian Vasile,Gaëtan Racic,Ugo Tanielian

2023-09-08

Abstract:In digital advertising, the selection of the optimal item (recommendation) and its best creative presentation (creative optimization) have traditionally been considered separate disciplines. However, both contribute significantly to user satisfaction, underpinning our assumption that it relies on both an item's relevance and its presentation, particularly in the case of visual creatives. In response, we introduce the task of {\itshape Generative Creative Optimization (GCO)}, which proposes the use of generative models for creative generation that incorporate user interests, and {\itshape AdBooster}, a model for personalized ad creatives based on the Stable Diffusion outpainting architecture. This model uniquely incorporates user interests both during fine-tuning and at generation time. To further improve AdBooster's performance, we also introduce an automated data augmentation pipeline. Through our experiments on simulated data, we validate AdBooster's effectiveness in generating more relevant creatives than default product images, showing its potential of enhancing user engagement.

Information Retrieval,Machine Learning

What problem does this paper attempt to address?

The paper attempts to address the issue of the separation between creative optimization and recommendation systems in digital advertising. Specifically, the authors argue that traditional recommendation systems focus solely on the quality of the recommended items, neglecting the impact of creative presentation (such as images and text descriptions) on user engagement. To tackle this challenge, the paper introduces the concept of Generative Creative Optimization (GCO) and proposes a new model based on the Stable Diffusion outpainting architecture—AdBooster.

The main objectives include:
1. **Introducing the GCO task**: Utilizing user interest signals to personalize the generation of ad creatives, making them fully aligned with the user interest signals from the recommendation system, thereby better matching user needs.
2. **Proposing the AdBooster model**: This is a model based on the Stable Diffusion outpainting architecture, capable of fine-tuning according to the user's interest distribution and considering user interests when generating creatives, ensuring that the generated creatives align with each user's preferences.
3. **Innovative data augmentation pipeline**: Proposing an automated data augmentation method to enhance the robustness and performance of the AdBooster model.
4. **Validating effectiveness**: Demonstrating through a series of experiments that the ad creatives generated by AdBooster are more relevant compared to default product images, thereby increasing user engagement.

Overall, the paper aims to significantly improve the effectiveness of digital advertising through personalized ad creative generation.