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

In industrial product quality assessment, it is essential to determinewhether a product is defect-free and further analyze the severity of anomality.To this end, accurate defect segmentation on images of products provides animportant functionality. In industrial inspection tasks, it is common tocapture abundant defect-free image samples but very limited anomalous ones.Therefore, it is critical to develop automatic and accurate defect segmentationsystems using only a small number of annotated anomalous training images. Thispaper tackles the challenging few-shot defect segmentation task with sufficientnormal (defect-free) training images but very few anomalous ones. We presenttwo effective regularization techniques via incorporating abundant defect-freeimages into the training of a UNet-like encoder-decoder defect segmentationnetwork. We first propose a Normal Background Regularization (NBR) loss whichis jointly minimized with the segmentation loss, enhancing the encoder networkto produce distinctive representations for normal regions. Secondly, wecrop/paste defective regions to the randomly selected normal images for dataaugmentation and propose a weighted binary cross-entropy loss to enhance thetraining by emphasizing more realistic crop-and-pasted augmented images basedon feature-level similarity comparison. Both techniques are implemented on anencoder-decoder segmentation network backboned by ResNet-34 for few-shot defectsegmentation. Extensive experiments are conducted on the recently releasedMVTec Anomaly Detection dataset with high-resolution industrial images. Underboth 1-shot and 5-shot defect segmentation settings, the proposed methodsignificantly outperforms several benchmarking methods.

Few-Shot Defect Segmentation Leveraging Abundant Normal Training Samples Through Normal Background Regularization and Crop-and-Paste Operation