{"ID":6620721,"CreatedAt":"2026-07-15T01:01:48.440468303Z","UpdatedAt":"2026-07-15T03:28:55.185153975Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2607.12866","arxiv_id":"2607.12866","title":"Statistical Non-linear Reconstruction Loss for Image Anomaly Detection","abstract":"Reconstruction-based methods are a cornerstone of unsupervised image anomaly detection, but they remain vulnerable to \\emph{outlier leakage}, where standard mean squared error (MSE) loss drives the model to faithfully reconstruct anomalous patterns. We propose a Non-linear Reconstruction Loss that applies a sigmoid-based squashing function to suppress high-magnitude features, preventing outliers from dominating optimization while preserving sensitivity to normal patterns. In addition, we introduce a statistical calibration scheme that selects the scaling factor $k$ from the confidence interval (CI) of the normal feature distribution, enabling data-driven control of the suppression strength. Our approach achieves competitive or superior anomaly detection performance compared to state-of-the-art methods, reaching 99.0\\% Image-AUROC and 97.3\\% Pixel-AUROC on MVTec-AD, and 95.3\\% Image-AUROC and 99.0\\% Pixel-AUROC on VisA. These results indicate that non-linear gradient suppression is an effective mechanism for mitigating outlier leakage and improving anomaly localization in unified industrial inspection settings. The implementation is available at https://github.com/mintii13/Statistical-Non-linear-Reconstruction-Loss.git.","short_abstract":"Reconstruction-based methods are a cornerstone of unsupervised image anomaly detection, but they remain vulnerable to \\emph{outlier leakage}, where standard mean squared error (MSE) loss drives the model to faithfully reconstruct anomalous patterns. We propose a Non-linear Reconstruction Loss that applies a sigmoid-bas...","url_abs":"https://arxiv.org/abs/2607.12866","url_pdf":"https://arxiv.org/pdf/2607.12866v1","authors":"[\"Nguyen Minh Tri\",\"Hoang Khuong Duy\",\"Huynh Cong Viet Ngu\"]","published":"2026-07-14T15:18:24Z","proceeding":"cs.CV","tasks":"[\"cs.CV\"]","methods":"[]","has_code":false,"code_links":[{"ID":614256,"CreatedAt":"2026-07-15T01:01:48.440468303Z","UpdatedAt":"2026-07-15T01:01:48.440468303Z","DeletedAt":null,"paper_id":6620721,"paper_url":"https://arxiv.org/abs/2607.12866","paper_title":"Statistical Non-linear Reconstruction Loss for Image Anomaly Detection","repo_url":"https://github.com/mintii13/Statistical-Non-linear-Reconstruction-Loss.git","is_official":false,"mentioned_in_paper":false,"mentioned_in_github":true,"github_stars":0}]}