Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

SARD: Segmentation-Aware Anomaly Synthesis via Region-Constrained Diffusion with Discriminative Mask Guidance

Yanshu Wang, Xichen Xu, Xiaoning Lei, Guoyang Xie

TL;DR

The paper tackles the challenge of generating spatially precise industrial anomalies for segmentation. It introduces SARD, a diffusion-based framework with Region-Constrained Diffusion to update only foreground regions and a Discriminative Mask Guidance discriminator for mask-aware supervision. Through extensive experiments on MVTec-AD and BTAD, SARD achieves substantial gains in pixel-level segmentation metrics and produces high-fidelity, mask-aligned anomalies. The work highlights the importance of foreground-background decoupling and region-specific adversarial feedback for effective data augmentation in industrial inspection.

Abstract

Synthesizing realistic and spatially precise anomalies is essential for enhancing the robustness of industrial anomaly detection systems. While recent diffusion-based methods have demonstrated strong capabilities in modeling complex defect patterns, they often struggle with spatial controllability and fail to maintain fine-grained regional fidelity. To overcome these limitations, we propose SARD (Segmentation-Aware anomaly synthesis via Region-constrained Diffusion with discriminative mask Guidance), a novel diffusion-based framework specifically designed for anomaly generation. Our approach introduces a Region-Constrained Diffusion (RCD) process that preserves the background by freezing it and selectively updating only the foreground anomaly regions during the reverse denoising phase, thereby effectively reducing background artifacts. Additionally, we incorporate a Discriminative Mask Guidance (DMG) module into the discriminator, enabling joint evaluation of both global realism and local anomaly fidelity, guided by pixel-level masks. Extensive experiments on the MVTec-AD and BTAD datasets show that SARD surpasses existing methods in segmentation accuracy and visual quality, setting a new state-of-the-art for pixel-level anomaly synthesis.

SARD: Segmentation-Aware Anomaly Synthesis via Region-Constrained Diffusion with Discriminative Mask Guidance

TL;DR

The paper tackles the challenge of generating spatially precise industrial anomalies for segmentation. It introduces SARD, a diffusion-based framework with Region-Constrained Diffusion to update only foreground regions and a Discriminative Mask Guidance discriminator for mask-aware supervision. Through extensive experiments on MVTec-AD and BTAD, SARD achieves substantial gains in pixel-level segmentation metrics and produces high-fidelity, mask-aligned anomalies. The work highlights the importance of foreground-background decoupling and region-specific adversarial feedback for effective data augmentation in industrial inspection.

Abstract

Synthesizing realistic and spatially precise anomalies is essential for enhancing the robustness of industrial anomaly detection systems. While recent diffusion-based methods have demonstrated strong capabilities in modeling complex defect patterns, they often struggle with spatial controllability and fail to maintain fine-grained regional fidelity. To overcome these limitations, we propose SARD (Segmentation-Aware anomaly synthesis via Region-constrained Diffusion with discriminative mask Guidance), a novel diffusion-based framework specifically designed for anomaly generation. Our approach introduces a Region-Constrained Diffusion (RCD) process that preserves the background by freezing it and selectively updating only the foreground anomaly regions during the reverse denoising phase, thereby effectively reducing background artifacts. Additionally, we incorporate a Discriminative Mask Guidance (DMG) module into the discriminator, enabling joint evaluation of both global realism and local anomaly fidelity, guided by pixel-level masks. Extensive experiments on the MVTec-AD and BTAD datasets show that SARD surpasses existing methods in segmentation accuracy and visual quality, setting a new state-of-the-art for pixel-level anomaly synthesis.

Paper Structure

This paper contains 10 sections, 9 equations, 2 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Proposed Method
  4. Region-Constrained Diffusion (RCD)
  5. Discriminative Mask Guidance (DMG)
  6. Experiments and Results
  7. Implementation Details
  8. Comparison Studies
  9. Ablation Studies
  10. Conclusion

Figures (2)

  • Figure 1: Overview of the proposed SARD framework. At each reverse diffusion step, the generator predicts a clean image, which is used to sample a posterior estimate. Region-Constrained Diffusion (RCD) fuses this sample with the noisy input, preserving background regions while updating only the anomaly area. A dual-branch discriminator enforces both global realism and mask-guided local fidelity.
  • Figure 2: Visual comparison on MVTec-AD. Each column is a method (CutPaste, DRAEM, GLASS, RealNet, DFMGAN, AnomalyDiffusion, DDGAN, SARD). Results highlight boundary precision and texture quality.