Implicit Concept Removal of Diffusion Models

TL;DR

This paper introduces Geom-Erasing, a geometric-driven method for removing unwanted implicit concepts like watermarks from diffusion models, and presents a new dataset for evaluating such concepts.

Contribution

The paper proposes a novel geometric-based approach for implicit concept removal and introduces the Implicit Concept Dataset for benchmarking.

Findings

Geom-Erasing effectively reduces implicit concept generation.

Achieves state-of-the-art results on I2P and ICD benchmarks.

Demonstrates robustness across different implicit concepts.

Abstract

Text-to-image (T2I) diffusion models often inadvertently generate unwanted concepts such as watermarks and unsafe images. These concepts, termed as the "implicit concepts", could be unintentionally learned during training and then be generated uncontrollably during inference. Existing removal methods still struggle to eliminate implicit concepts primarily due to their dependency on the model's ability to recognize concepts it actually can not discern. To address this, we utilize the intrinsic geometric characteristics of implicit concepts and present the Geom-Erasing, a novel concept removal method based on the geometric-driven control. Specifically, once an unwanted implicit concept is identified, we integrate the existence and geometric information of the concept into the text prompts with the help of an accessible classifier or detector model. Subsequently, the model is optimized to identify and disentangle this information, which is then adopted as negative prompts during generation. Moreover, we introduce the Implicit Concept Dataset (ICD), a novel image-text dataset imbued with three typical implicit concepts (i.e., QR codes, watermarks, and text), reflecting real-life situations where implicit concepts are easily injected. Geom-Erasing effectively mitigates the generation of implicit concepts, achieving the state-of-the-art results on the Inappropriate Image Prompts (I2P) and our challenging Implicit Concept Dataset (ICD) benchmarks.