Automatic Segmentation of Early Triassic Vertebrate Fossil CT Scans: Reducing Human Annotation Time through Deep Learning DOI Creative Commons
Espen M. Knutsen, Dmitry A. Konovalov

Research Square (Research Square), Journal Year: 2024, Volume and Issue: unknown

Published: May 21, 2024

Abstract Recent developments in Deep Learning have opened the possibility for automated segmentation of large and highly detailed CT scan datasets fossil material. However, previous methodologies required amounts training data to reliably extract complex skeletal structures. Here we present a method obtain high-fidelity 3D models fossils digitally extracted from surrounding rock, model with less than 1%-2% total dataset. This workflow has capacity revolutionise use significantly reduce processing time such boost availability segmented CT-scanned material future research outputs. Our final Unet achieved validation Dice similarity 0.96.

Language: Английский

Not all scans are equal: X-ray tomography image quality evaluation DOI
Anton du Plessis, Muofhe Tshibalanganda, S. Roux

et al.

Materials Today Communications, Journal Year: 2019, Volume and Issue: 22, P. 100792 - 100792

Published: Nov. 22, 2019

Language: Английский

Citations

62
Multimodal Defect Imaging of Pure Tungsten Components Fabricated via Electron Beam Powder Bed Fusion DOI Creative Commons

H.M. Zhang,

Paul Carriere,

Dan Schneberk

et al.

Journal of Materials Engineering and Performance, Journal Year: 2025, Volume and Issue: unknown

Published: March 7, 2025

Abstract The utilization of additive manufacturing (AM) techniques for refractory materials in high-temperature environments has significantly expanded because the ability to fabricate geometrically complex components. Electron beam powder bed fusion (EB-PBF), which provides lower residual stress, a cleaner vacuum environment, and better efficiency high melting point, is one best-suited AM methods produce advanced However, property variation attributed heterogeneous microstructure process-induced defects hindered widespread adoption EB-PBF-produced material like tungsten. While numerous in-situ monitoring defect detection have been demonstrated EB-PBF, workflow that compares evaluates abnormalities from different imaging perspectives still limited. This study examines feature-embedded tungsten component manufactured via EB-PBF process demonstrate capabilities multimodal workflow. predefined are evaluated by harnessing various techniques, including electron imaging, layerwise near-infrared (NIR) post-build high-energy x-ray computed tomography (CT), conventional destructive metallography. results highlight strengths limitations distinctive concerning specific types, sizes, conditions. It was found can provide more abnormal while maintaining higher measuring accuracy, against metallography this case study, compared with NIR CT techniques.

Language: Английский

Citations

0
Accelerating segmentation of fossil CT scans through Deep Learning DOI Creative Commons
Espen M. Knutsen, Dmitry A. Konovalov

Scientific Reports, Journal Year: 2024, Volume and Issue: 14(1)

Published: Sept. 9, 2024

Language: Английский

Citations

2
Characterization of self-piercing rivet joints using X-ray computed tomography DOI Creative Commons
Papangkorn Jessadatavornwong,

Garret Huff,

Amanda Freis

et al.

Tomography of Materials and Structures, Journal Year: 2023, Volume and Issue: 2, P. 100010 - 100010

Published: June 1, 2023

Self-piercing rivet (SPR) joining is a process that has been adopted in the automotive industry, and it important to be able characterize SPR joints non-destructively. X-ray computed tomography (CT) allows visualizing internal structure of sample whilst preserving workpiece. In this work, application CT joint characterization investigated. Many features riveted can observed. This includes boundary between sheet interfaces position where substrates meet, which are quality assessment. case similar materials, some these become difficult observe but presence adhesive sheets enables more easily Furthermore, substrate fracture cracks, radial cracks asymmetry detected due difference greyscale surrounding materials. It also revealed image resolution plays an role defect detectability technique.

Language: Английский

Citations

3
Testing the impact of two key scan parameters on the quality and repeatability of measurements from CT scan data DOI Creative Commons
Rosie L. Oakes,

Morgan Hill Chase,

Mark E. Siddall

et al.

Palaeontologia Electronica, Journal Year: 2020, Volume and Issue: unknown

Published: Jan. 1, 2020

Computed tomographic (CT) scanning is becoming a popular research tool across earth and life sciences.However, despite its prominence, there have not been systematic investigations into how CT scan parameters affect data quality reproducibility.Here we conduct two sets of trials to test exposure time, the number x-ray radiographs averaged per view, overall time data, assessed using signal contrast noise ratios repeatability measurements derived from these in this case calculated volume pteropod shells.We find that ratio shell increase variability decrease with increasing time.However, benefits increased diminish considerably at times 50 minutes or more.Furthermore, as increases, scans are greater risk being affected by sample movement, which can make unusable.By balancing image 50-minute be comparable to, better than, collected 75-minute scan.By selecting rather than scan, collection between 66 75%, maximizing both quantity collected.

Language: Английский

Citations

6
Automatic Segmentation of Early Triassic Vertebrate Fossil CT Scans: Reducing Human Annotation Time through Deep Learning DOI Creative Commons
Espen M. Knutsen, Dmitry A. Konovalov

Research Square (Research Square), Journal Year: 2024, Volume and Issue: unknown

Published: May 21, 2024

Abstract Recent developments in Deep Learning have opened the possibility for automated segmentation of large and highly detailed CT scan datasets fossil material. However, previous methodologies required amounts training data to reliably extract complex skeletal structures. Here we present a method obtain high-fidelity 3D models fossils digitally extracted from surrounding rock, model with less than 1%-2% total dataset. This workflow has capacity revolutionise use significantly reduce processing time such boost availability segmented CT-scanned material future research outputs. Our final Unet achieved validation Dice similarity 0.96.

Language: Английский

Citations

0