2022
DOI: 10.1007/s40192-022-00272-5
|View full text |Cite
|
Sign up to set email alerts
|

Image Segmentation for Defect Analysis in Laser Powder Bed Fusion: Deep Data Mining of X-Ray Photography from Recent Literature

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1
1

Citation Types

0
5
0

Year Published

2023
2023
2024
2024

Publication Types

Select...
5
2

Relationship

0
7

Authors

Journals

citations
Cited by 7 publications
(5 citation statements)
references
References 33 publications
0
5
0
Order By: Relevance
“…As shown in Table 2 below, IoU and BF-score are high and close to 1 for both datasets 1 and 2 for run 1, which means that this method successfully segments out the keyhole region in test datasets. Considering IoU, this method is on the same level as other segmentation tools proposed in previous works, and a high BF-score further validates the segmentation accuracy on the boundary [14]. The same tool is later tested using cross-validation, with three more runs trained and tested as Table 3 below.…”
Section: Testing and Performance Matricesmentioning
confidence: 83%
See 1 more Smart Citation
“…As shown in Table 2 below, IoU and BF-score are high and close to 1 for both datasets 1 and 2 for run 1, which means that this method successfully segments out the keyhole region in test datasets. Considering IoU, this method is on the same level as other segmentation tools proposed in previous works, and a high BF-score further validates the segmentation accuracy on the boundary [14]. The same tool is later tested using cross-validation, with three more runs trained and tested as Table 3 below.…”
Section: Testing and Performance Matricesmentioning
confidence: 83%
“…More specifically, sufficient data for machine learning-based pore prediction for many types of LPBF systems, conditions, and alloys, will require very large efforts to label keyhole morphologies unless accessible automatic tools that can segment the keyholes are developed. Several automatic tools have recently been explored in previous works: Pyeon et al developed a non-machine learning-based semi-automatic keyhole region extraction tool [13], and Zhang et al tested several semantic segmentation and object detection models and compared the performances of extracting keyholes and pores at the same time [14]. However, the filter developed by Pyeon et al was only tested with clean images without metal powder, and models tested by Zhang et al segment both keyholes and pores in the same classification.…”
Section: Introductionmentioning
confidence: 99%
“…As discussed in Section 4.2, the detachment-based SOTA involves using image segmentation to determine when the melt pool (MP) divides into separate volumes. Image segmentation has garnered significant attention in additive manufacturing (AM) research, leading to various successful models tailored to specific AM-related segmentation problems [31]. On the contrary, the Segment Anything Model (SAM) [30], developed by Meta, stands out due to its zero-shot learning capabilities, which allow it to operate across new datasets without further training or finetuning.…”
Section: Appendix Amentioning
confidence: 99%
“…As discussed in Section0, the detachment-based SOTA involves using image segmentation to determine when the melt pool (MP) divides into separate volumes. Image segmentation has garnered significant attention in additive manufacturing (AM) research, leading to various successful models tailored to specific AM-related segmentation problems [31]. On the contrary, the Segment Anything Model (SAM) [30], developed by Meta, stands out due to its zero-shot learning capabilities, which allow it to operate across new datasets without further training or finetuning.…”
Section: Fundingmentioning
confidence: 99%