Mitigating Memorization in Text-to-Image Diffusion via Region-Aware Prompt Augmentation and Multimodal Copy Detection
This addresses copyright and privacy risks in text-to-image generation, but it is incremental as it builds on existing prompt perturbation methods.
The paper tackles the problem of text-to-image diffusion models memorizing and reproducing training images, which poses copyright and privacy risks, by introducing Region-Aware Prompt Augmentation (RAPTA) and Attention-Driven Multimodal Copy Detection (ADMCD), with results showing RAPTA reduces overfitting while maintaining synthesis quality and ADMCD reliably detects copying, outperforming single-modal metrics.
State-of-the-art text-to-image diffusion models can produce impressive visuals but may memorize and reproduce training images, creating copyright and privacy risks. Existing prompt perturbations applied at inference time, such as random token insertion or embedding noise, may lower copying but often harm image-prompt alignment and overall fidelity. To address this, we introduce two complementary methods. First, Region-Aware Prompt Augmentation (RAPTA) uses an object detector to find salient regions and turn them into semantically grounded prompt variants, which are randomly sampled during training to increase diversity, while maintaining semantic alignment. Second, Attention-Driven Multimodal Copy Detection (ADMCD) aggregates local patch, global semantic, and texture cues with a lightweight transformer to produce a fused representation, and applies simple thresholded decision rules to detect copying without training with large annotated datasets. Experiments show that RAPTA reduces overfitting while maintaining high synthesis quality, and that ADMCD reliably detects copying, outperforming single-modal metrics.