Engineering and Technology

Deep learning corrosion detection with confidence

W. Nash, L. Zheng, et al.

Discover how researchers Will Nash, Liang Zheng, and Nick Birbilis have developed a groundbreaking deep learning model for pixel-level corrosion segmentation, enhancing economic safety with confidence estimates. Their innovative approach outperforms existing solutions and brings new insights into decision-making processes!

00:00

Playback language: English

Index

Abstract

Corrosion detection is crucial due to its significant economic impact. This paper presents a deep learning model for pixel-level corrosion segmentation, incorporating Bayesian methods to provide confidence estimates. Experiments on a newly collected dataset show promising results, exceeding existing state-of-the-art accuracy while offering uncertainty measures to improve decision-making.

Publisher

npj Materials Degradation

Published On

Mar 31, 2022

Authors

Will Nash, Liang Zheng, Nick Birbilis

DOI

https://doi.org/10.1038/s41529-022-00232-6