CAT: Contrastive Adapter Training for Personalized Image Generation
This addresses a specific issue in personalized image generation for users of diffusion models, but it is incremental as it builds on existing adapter methods.
The paper tackles the problem of adapter training in diffusion models leading to loss of diversity and corruption of prior knowledge, and presents Contrastive Adapter Training (CAT) with a new loss function that improves performance, as shown through qualitative and quantitative comparisons.
The emergence of various adapters, including Low-Rank Adaptation (LoRA) applied from the field of natural language processing, has allowed diffusion models to personalize image generation at a low cost. However, due to the various challenges including limited datasets and shortage of regularization and computation resources, adapter training often results in unsatisfactory outcomes, leading to the corruption of the backbone model's prior knowledge. One of the well known phenomena is the loss of diversity in object generation, especially within the same class which leads to generating almost identical objects with minor variations. This poses challenges in generation capabilities. To solve this issue, we present Contrastive Adapter Training (CAT), a simple yet effective strategy to enhance adapter training through the application of CAT loss. Our approach facilitates the preservation of the base model's original knowledge when the model initiates adapters. Furthermore, we introduce the Knowledge Preservation Score (KPS) to evaluate CAT's ability to keep the former information. We qualitatively and quantitatively compare CAT's improvement. Finally, we mention the possibility of CAT in the aspects of multi-concept adapter and optimization.