MoA: Mixture-of-Attention for Subject-Context Disentanglement in Personalized Image Generation
This addresses the need for better control over personalized image generation for users, though it appears incremental as it builds on existing diffusion models and Mixture-of-Experts mechanisms.
The paper tackles the problem of subject-context entanglement in personalized text-to-image generation by introducing Mixture-of-Attention (MoA), which uses two attention pathways to separate personalized and generic content, resulting in high-quality images with diverse compositions and interactions.
We introduce a new architecture for personalization of text-to-image diffusion models, coined Mixture-of-Attention (MoA). Inspired by the Mixture-of-Experts mechanism utilized in large language models (LLMs), MoA distributes the generation workload between two attention pathways: a personalized branch and a non-personalized prior branch. MoA is designed to retain the original model's prior by fixing its attention layers in the prior branch, while minimally intervening in the generation process with the personalized branch that learns to embed subjects in the layout and context generated by the prior branch. A novel routing mechanism manages the distribution of pixels in each layer across these branches to optimize the blend of personalized and generic content creation. Once trained, MoA facilitates the creation of high-quality, personalized images featuring multiple subjects with compositions and interactions as diverse as those generated by the original model. Crucially, MoA enhances the distinction between the model's pre-existing capability and the newly augmented personalized intervention, thereby offering a more disentangled subject-context control that was previously unattainable. Project page: https://snap-research.github.io/mixture-of-attention