tl;dr Methods like Dreambooth lead to substantial drift in previously learned representations during the finetuning process even when adapting to as few as five images.
Recent advances in diffusion models have significantly enhanced image generation capabilities. However, customizing these models with new classes often leads to unintended consequences that compromise their reliability. We introduce the concept of open-world forgetting to emphasize the vast scope of these unintended alterations, contrasting it with the well-studied closed-world forgetting, which is measurable by evaluating performance on a limited set of classes or skills. Our research presents the first comprehensive investigation into open-world forgetting in diffusion models, focusing on semantic and appearance drift of representations. We utilize zero-shot classification to analyze semantic drift, revealing that even minor model adaptations lead to unpredictable shifts affecting areas far beyond newly introduced concepts, with dramatic drops in zero-shot classification of up to 60%. Additionally, we observe significant changes in texture and color of generated content when analyzing appearance drift. To address these issues, we propose a mitigation strategy based on functional regularization, designed to preserve original capabilities while accommodating new concepts. Our study aims to raise awareness of unintended changes due to model customization and advocates for the analysis of open-world forgetting in future research on model customization and finetuning methods. Furthermore, we provide insights for developing more robust adaptation methodologies.
@article{laria2024forgetting,
title={Assessing Open-world Forgetting in Generative Image Model Customization},
author={H\'ector Laria and Alex Gomez-Villa and Imad Eddine Marouf and Kai Wang and Bogdan Raducanu and Joost van de Weijer},
booktitle={arXiv preprint arxiv:2410.14159},
year={2024}
}