Built independently by an author, for readers. Read the story and support ChapterPal
ChapterPal Logo
Log inSign up

keyword

patch-level denoising strategy

A patch-level denoising strategy is a data-filtering and noise-reduction technique in computer vision and machine learning that detects, attenuates, or removes corrupted or anomalous information at the granularity of localized image sub-regions rather than discarding entire images. Instead of relying on sample-level filtering that discards an entire image when only a small portion is corrupted or mislabeled, this approach evaluates feature representations across individual patches and assigns local noise or outlier scores. By selectively eliminating or downweighting corrupted patches while preserving the valid, normal regions of a noisy sample, the strategy prevents corrupted data from distorting decision boundaries, improves memory bank and coreset quality, and enhances model robustness in tasks such as unsupervised anomaly detection.

1 item

SoftPatch: Unsupervised Anomaly Detection with Noisy Data

SoftPatch: Unsupervised Anomaly Detection with Noisy Data

Xi Jiang, Jianlin Liu, Jinbao Wang, Qiang Nie, Kai Wu, Yong Liu, Chengjie Wang, Feng Zheng

OrganizationsDepartment of Computer Science and EngineeringSouthern University of Science and TechnologyTencent

Why you should read this

Proposes SoftPatch, a patch-level denoising and memory re-weighting method that prevents defective training samples from distorting decision boundaries in real-world unsupervised anomaly detection.

Although mainstream unsupervised anomaly detection (AD) algorithms perform well in academic datasets, their performance is limited in practical application due to the ideal experimental setting of clean training data. Training with noisy data is an inevitable problem in real-world anomaly detection but is seldom discussed. This paper considers label-level noise in image sensory anomaly detection for the first time. To solve this problem, we proposed a memory-based unsupervised AD method, SoftPatch, which efficiently denoises the data at the patch level. Noise discriminators are utilized to generate outlier scores for patch-level noise elimination before coreset construction. The scores are then stored in the memory bank to soften the anomaly detection boundary. Compared with existing methods, SoftPatch maintains a strong modeling ability of normal data and alleviates the overconfidence problem in coreset. Comprehensive experiments in various noise scenes demonstrate that SoftPatch outperforms the state-of-the-art AD methods on the MVTecAD and BTAD benchmarks and is comparable to those methods under the setting without noise.

Added

2026-09-26