OmiAD: One-Step Adaptive Masked Diffusion Model for Multi-class Anomaly Detection via Adversarial Distillation

Yaoxuan Feng, Wenchao Chen, Yuxin Li, Bo Chen, Yubiao Wang, Zixuan Zhao, Hongwei Liu, Mingyuan Zhou
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:16604-16633, 2025.

Abstract

Diffusion models have demonstrated outstanding performance in industrial anomaly detection. However, their iterative denoising nature results in slow inference speed, limiting their practicality for real-time industrial deployment. To address this challenge, we propose OmiAD, a one-step masked diffusion model for multi-class anomaly detection, derived from a well-designed multi-step Adaptive Masked Diffusion Model (AMDM) and compressed using Adversarial Score Distillation (ASD). OmiAD first introduces AMDM, equipped with an adaptive masking strategy that dynamically adjusts masking patterns based on noise levels and encourages the model to reconstruct anomalies as normal counterparts by leveraging broader context, to reduce the pixel-level shortcut reliance. Then, ASD is developed to compress the multi-step diffusion process into a single-step generator by score distillation and incorporating a shared-weight discriminator effectively reusing parameters while significantly improving both inference efficiency and detection performance. The effectiveness of OmiAD is validated on four diverse datasets, achieving state-of-the-art performance across seven metrics while delivering a remarkable inference speedup.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-feng25b, title = {{O}mi{AD}: One-Step Adaptive Masked Diffusion Model for Multi-class Anomaly Detection via Adversarial Distillation}, author = {Feng, Yaoxuan and Chen, Wenchao and Li, Yuxin and Chen, Bo and Wang, Yubiao and Zhao, Zixuan and Liu, Hongwei and Zhou, Mingyuan}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {16604--16633}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/feng25b/feng25b.pdf}, url = {https://proceedings.mlr.press/v267/feng25b.html}, abstract = {Diffusion models have demonstrated outstanding performance in industrial anomaly detection. However, their iterative denoising nature results in slow inference speed, limiting their practicality for real-time industrial deployment. To address this challenge, we propose OmiAD, a one-step masked diffusion model for multi-class anomaly detection, derived from a well-designed multi-step Adaptive Masked Diffusion Model (AMDM) and compressed using Adversarial Score Distillation (ASD). OmiAD first introduces AMDM, equipped with an adaptive masking strategy that dynamically adjusts masking patterns based on noise levels and encourages the model to reconstruct anomalies as normal counterparts by leveraging broader context, to reduce the pixel-level shortcut reliance. Then, ASD is developed to compress the multi-step diffusion process into a single-step generator by score distillation and incorporating a shared-weight discriminator effectively reusing parameters while significantly improving both inference efficiency and detection performance. The effectiveness of OmiAD is validated on four diverse datasets, achieving state-of-the-art performance across seven metrics while delivering a remarkable inference speedup.} }
Endnote
%0 Conference Paper %T OmiAD: One-Step Adaptive Masked Diffusion Model for Multi-class Anomaly Detection via Adversarial Distillation %A Yaoxuan Feng %A Wenchao Chen %A Yuxin Li %A Bo Chen %A Yubiao Wang %A Zixuan Zhao %A Hongwei Liu %A Mingyuan Zhou %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-feng25b %I PMLR %P 16604--16633 %U https://proceedings.mlr.press/v267/feng25b.html %V 267 %X Diffusion models have demonstrated outstanding performance in industrial anomaly detection. However, their iterative denoising nature results in slow inference speed, limiting their practicality for real-time industrial deployment. To address this challenge, we propose OmiAD, a one-step masked diffusion model for multi-class anomaly detection, derived from a well-designed multi-step Adaptive Masked Diffusion Model (AMDM) and compressed using Adversarial Score Distillation (ASD). OmiAD first introduces AMDM, equipped with an adaptive masking strategy that dynamically adjusts masking patterns based on noise levels and encourages the model to reconstruct anomalies as normal counterparts by leveraging broader context, to reduce the pixel-level shortcut reliance. Then, ASD is developed to compress the multi-step diffusion process into a single-step generator by score distillation and incorporating a shared-weight discriminator effectively reusing parameters while significantly improving both inference efficiency and detection performance. The effectiveness of OmiAD is validated on four diverse datasets, achieving state-of-the-art performance across seven metrics while delivering a remarkable inference speedup.
APA
Feng, Y., Chen, W., Li, Y., Chen, B., Wang, Y., Zhao, Z., Liu, H. & Zhou, M.. (2025). OmiAD: One-Step Adaptive Masked Diffusion Model for Multi-class Anomaly Detection via Adversarial Distillation. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:16604-16633 Available from https://proceedings.mlr.press/v267/feng25b.html.

Related Material