{"ID":5937156,"CreatedAt":"2026-07-07T03:14:33.014478982Z","UpdatedAt":"2026-07-09T10:17:55.271822851Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2607.04894","arxiv_id":"2607.04894","title":"ProCon: Projection-Consistency Memory for Training-Free Anomaly Detection","abstract":"Memory-based anomaly detection is attractive because it localizes defects from normal images without training a decoder or synthesizing pseudo anomalies. However, most memory methods still use the memory bank as a nearest-neighbor lookup table: a test patch is treated as normal if it has one nearby normal anchor. This hard retrieval view is vulnerable to false-normal matches and does not test whether the patch is consistently supported by a local normal neighborhood. We propose ProCon, a training-free framework that turns memory retrieval into decoder-free reconstruction. ProCon softly projects each test patch onto nearby normal memory vectors and uses the projection residual as anomaly evidence. To stabilize this residual, it constructs seed-perturbed layer-wise memories, aggregates bank residuals by a median, and fuses depth-specific residual maps by layer consensus. ProCon requires no decoder training, backbone fine-tuning, learned fusion weights, or pseudo-anomaly supervision. Across MVTec-AD, VisA, and Real-IAD under the single-category evaluation protocol, ProCon achieves strong image- and pixel-level performance under seven standard metrics, including image AUROC scores of 99.8%, 99.2%, and 93.2%, respectively. Ablations show that the gains come from replacing hard retrieval with soft normal projection and stabilizing the residuals through memory and depth consensus. The code is available at https://github.com/jw-chae/Procon","short_abstract":"Memory-based anomaly detection is attractive because it localizes defects from normal images without training a decoder or synthesizing pseudo anomalies. However, most memory methods still use the memory bank as a nearest-neighbor lookup table: a test patch is treated as normal if it has one nearby normal anchor. This...","url_abs":"https://arxiv.org/abs/2607.04894","url_pdf":"https://arxiv.org/pdf/2607.04894v1","authors":"[\"Joongwon Chae\",\"Lihui Luo\",\"Yang Liu\",\"Dongmei Yu\",\"Peiwu Qin\",\"Runming Wang\",\"Ilmoon Chae\"]","published":"2026-07-06T10:23:24Z","proceeding":"cs.CV","tasks":"[\"cs.CV\"]","methods":"[]","has_code":false,"code_links":[{"ID":613961,"CreatedAt":"2026-07-07T03:14:33.014478982Z","UpdatedAt":"2026-07-07T03:14:33.014478982Z","DeletedAt":null,"paper_id":5937156,"paper_url":"https://arxiv.org/abs/2607.04894","paper_title":"ProCon: Projection-Consistency Memory for Training-Free Anomaly Detection","repo_url":"https://github.com/jw-chae/Procon","is_official":false,"mentioned_in_paper":false,"mentioned_in_github":true,"github_stars":0}]}