Anomaly-Preference Image Generation

TL;DR

This paper introduces a novel anomaly image generation method that improves realism and diversity by preference learning and dynamic capacity allocation, outperforming existing techniques.

Contribution

It proposes Anomaly Preference Optimization with implicit preference alignment and a Time-Aware Capacity Allocation module for enhanced anomaly synthesis.

Findings

Achieves state-of-the-art realism and diversity in anomaly generation.

Outperforms existing baselines in experimental evaluations.

Provides a hierarchical sampling strategy for controlled generation.

Abstract

Synthesizing realistic and diverse anomalous samples from limited data is vital for robust model generalization. However, existing methods struggle to reconcile fidelity and diversity, often hampered by distribution misalignment and overfitting, respectively.To mitigate this, we introduce Anomaly Preference Optimization,a novel paradigm that reformulates anomaly generation as a preference learning problem.Central to our approach is an implicit preference alignment mechanism that leverages real anomalies as positive references, deriving optimization signals directly from denoising trajectory deviations without requiring costly human annotation. Furthermore, we propose a Time-Aware Capacity Allocation module that dynamically distributes model capacity along the diffusion timeline,prioritizing structural diversity during highnoise phases while enhancing fine-grained fidelity in low-noise stages. During inference, a hierarchical sampling strategy modulates the coherencealignment trade-off, enabling precise control over generation. Extensive experiments demonstrate that significantly outperforms existing baselines,achieving state-of-the-art performance in both realism and diversity.