Segmentation of multiple features in carbon fiber reinforced polymers using a convolutional neural network

Weinberger, Patrick; Plank, Bernhard; Froehler, Bernhard; Kastner, Johann; Heinzl, Christoph

doi:10.1117/12.2612895

Cited by 1 publication

(2 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In segmentation, labeled images are often used in a supervised manner to train a neural network for the task at hand [4]. The generation of labeled data is one of the most time-consuming tasks in the field of supervised learning.…”

Section: Related Workmentioning

confidence: 99%

“…Accurate and automated segmentation of XCT data is of paramount importance to efficiently identify and characterize these features, facilitating reliable defect detection and quality assurance in industrial components. Traditional supervised segmentation methods have proven effective in various medical imaging applications [3], but their reliance on extensive labeled datasets poses significant challenges when applied to the diverse components and materials encountered in industrial XCT [4]. The time-consuming and expensive process of obtaining precise annotations for industrial XCT data often restricts the scalability and practicality of supervised approaches in an industrial context.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Unsupervised Segmentation of Industrial X-ray Computed Tomography Data with the Segment Anything Model

Weinberger,

Schwarz,

Fröhler

et al. 2024

Preprint

Self Cite

View full text Add to dashboard Cite

Industrial X-ray computed tomography (XCT) is a crucial non-destructive testing method for quality control in various branches of industry. Accurate segmentation of XCT data aids in defect identification and material characterization. In this paper we present our results of applying the Segment Anything Model (SAM) in this context. SAM, an unsupervised approach that automates segmentation without manual annotations, combines deep convolutional neural networks and generative adversarial networks. We used SAM on diverse industrial XCT data, and our evaluation demonstrates SAM’s competitive segmentation accuracy and efficiency. SAM’s unsupervised nature and adaptability offer cost-effective and reliable quality assurance solutions in industrial settings, enhancing product quality and inspection processes.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%