English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Reconstruction-based methods are widely explored in industrial visual anomalydetection. Such methods commonly require the model to well reconstruct thenormal patterns but fail in the anomalies, and thus the anomalies can bedetected by evaluating the reconstruction errors. However, in practice, it'susually difficult to control the generalization boundary of the model. Themodel with an overly strong generalization capability can even well reconstructthe abnormal regions, making them less distinguishable, while the model with apoor generalization capability can not reconstruct those changeablehigh-frequency components in the normal regions, which ultimately leads tofalse positives. To tackle the above issue, we propose a new reconstructionnetwork where we reconstruct the original RGB image from its gray value edges(EdgRec). Specifically, this is achieved by an UNet-type denoising autoencoderwith skip connections. The input edge and skip connections can well preservethe high-frequency information in the original image. Meanwhile, the proposedrestoration task can force the network to memorize the normal low-frequency andcolor information. Besides, the denoising design can prevent the model fromdirectly copying the original high-frequent components. To evaluate theanomalies, we further propose a new interpretable hand-crafted evaluationfunction that considers both the color and gradient differences. Our methodachieves competitive results on the challenging benchmark MVTec AD (97.8\% fordetection and 97.7\% for localization, AUROC). In addition, we conductexperiments on the MVTec 3D-AD dataset and show convincing results using RGBimages only. Our code will be available athttps://github.com/liutongkun/EdgRec.

Reconstruction from edge image combined with color and gradient difference for industrial surface anomaly detection