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

Segmentation-based, two-stage neural network has shown excellent results inthe surface defect detection, enabling the network to learn from a relativelysmall number of samples. In this work, we introduce end-to-end training of thetwo-stage network together with several extensions to the training process,which reduce the amount of training time and improve the results on the surfacedefect detection tasks. To enable end-to-end training we carefully balance thecontributions of both the segmentation and the classification loss throughoutthe learning. We adjust the gradient flow from the classification into thesegmentation network in order to prevent the unstable features from corruptingthe learning. As an additional extension to the learning, we proposefrequency-of-use sampling scheme of negative samples to address the issue ofover- and under-sampling of images during the training, while we employ thedistance transform algorithm on the region-based segmentation masks as weightsfor positive pixels, giving greater importance to areas with higher probabilityof presence of defect without requiring a detailed annotation. We demonstratethe performance of the end-to-end training scheme and the proposed extensionson three defect detection datasets - DAGM, KolektorSDD and Severstal Steeldefect dataset - where we show state-of-the-art results. On the DAGM and theKolektorSDD we demonstrate 100\% detection rate, therefore completely solvingthe datasets. Additional ablation study performed on all three datasetsquantitatively demonstrates the contribution to the overall result improvementsfor each of the proposed extensions.

End-to-end training of a two-stage neural network for defect detection