Pith. sign in

REVIEW 5 major objections 6 minor 1 cited by

TransMatch: A Transfer-Learning Framework for Defect Detection in Laser Powder Bed Fusion Additive Manufacturing

T0 review · 5 major / 6 minor · reviewed 2026-08-05 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read TransMatch classifies LPBF surface defects — Crack, Pinhole, Hole, Spatter — at 98.91% accuracy from just 34 manually labeled FE-SEM images by combining four rounds of pseudo-labeling with transfer learning.

desk verdict Standard pseudo-labeling plus transfer learning applied to LPBF defect detection; the reported 98.91% accuracy is unsupported by inconsistent dataset counts and an unestablished test split. read the letter →

arxiv 2509.01754 v1 pith:AWOMU4HH submitted 2025-09-01 cs.CV physics.comp-ph

Mohsen Asghari Ilani , Yaser Mike Banad This is my paper
classification cs.CVphysics.comp-ph
keywords laserpowderbedfusiondefectdetectiontransferlearningsemi-supervisedfew-shotpseudo-labelingconvolutionalneuralnetworksFE-SEMimaging
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

TransMatch is a recipe for building a defect classifier for laser powder bed fusion (LPBF) parts from very little labeled data. The authors claim that a CNN trained on crops from just 34 manually annotated FE-SEM images can be extended, through four rounds of pseudo-labeling that admit only predictions above 50% confidence, to classify four surface-defect types — Crack, Pinhole, Hole, Spatter — with 98.91% accuracy and 0.0188 loss on a test set of 829 unlabeled images. Transfer learning supplies the initial feature representation, so the few labeled examples suffice to separate the defect classes. If the numbers hold, the framework would substantially reduce the annotation burden in additive-manufacturing quality control, where defect-labeled imagery is scarce and expensive to produce. The paper's own confusion analysis says residual errors concentrate on visually overlapping or co-occurring defects, such as cracks confused with pinholes or spatter.

What carries the argument

The load-bearing mechanism is iterative, thresholded pseudo-labeling fused with transfer learning. A seed CNN, trained on human-labeled defect crops, scores unlabeled images; predictions whose softmax confidence exceeds 50% are folded into the training set; the network is retrained on the expanded set; and the cycle repeats for four rounds. The confidence threshold is the control that trades pseudo-label quantity against label noise. Transfer learning supplies the inductive bias that makes few-shot separation possible: a feature extractor pre-trained on a large general corpus is fine-tuned on defect imagery, so the four defect classes (the novel classes of the few-shot setup) can be classifi

What would settle it

Re-run the TransMatch pipeline with an image-level split, assigning every crop from each of the 232 parent FE-SEM images to exactly one partition so that no test crop shares a parent image with any training or pseudo-labeled crop. If accuracy drops materially below 98.91%, the headline number was inflated by parent-image leakage.

Watch

Extended reading notes

Core claim

The paper's central claim is that TransMatch — semi-supervised few-shot learning built from thresholded pseudo-labeling plus transfer learning — detects LPBF surface defects at near-supervised accuracy from only 34 manually annotated FE-SEM images. A CNN trained on 6,742 preprocessed crops from those images pseudo-labels 7,455 unlabeled images, admitting only predictions above a 50% softmax-confidence threshold, retraining on the union, and repeating for four rounds, with features transferred from a network pre-trained on a large general corpus. On 829 held-out unlabeled images it reports 98.91% accuracy with 0.0188 loss, per-class precision and recall between 0.94 and 1.00, and F1 scores of

Load-bearing premise

The reported 98.91% test accuracy assumes that the 829 test images were held out cleanly from every image used in training and pseudo-labeling; the paper never documents how the 232 parent FE-SEM images were divided into labeled, unlabeled, and test sets, so test patches sharing a parent image with training patches could inflate the score.

Editorial extensions

If this is right

  • If the reported accuracy is real, defect classifiers for LPBF and other metal additive processes could be bootstrapped from tens of labeled images instead of thousands, cutting the dominant cost of building quality-control datasets.
  • The pseudo-labeling schedule is architecture-agnostic and additive: it can be attached to an already-trained CNN without redesign, so existing defect detectors could be upgraded rather than replaced.
  • Per-class F1 scores of 0.97–1.00 indicate that the four defect classes are separable in ex-situ FE-SEM imagery at a level useful for automated screening, not just for research benchmarks.
  • Because the supervised (99%) and semi-supervised (98.91%) pipelines finish nearly equal, the framework's practical value lies in flexibility and scalability under scarce labels rather than in a large accuracy gain over its own supervised baseline.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The near-identical supervised and semi-supervised test numbers suggest pseudo-labeling added coverage rather than accuracy; an ablation study withholding the four pseudo-labeling rounds would reveal the true marginal contribution of the unlabeled pool.
  • A 50% confidence threshold sits only slightly above chance for a four-class softmax, so many low-confidence pseudo-labels enter the training set; entropy-based or uncertainty-weighted thresholds could buy cleaner labels at a small recall cost.
  • Nothing in the pipeline is specific to FE-SEM or to these four classes, so the recipe should transfer to optical microscopy, X-ray CT, and other defect taxonomies — a cheap way to test its generality.
  • The decisive industrial test is image-level generalization: if pseudo-labeled crops share parent FE-SEM images with test crops, the quoted numbers inflate; a strict per-image split would settle it.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 6 minor

Summary. The paper proposes TransMatch, a transfer-learning framework combined with semi-supervised few-shot learning for classifying surface defects (Crack, Pinhole, Hole, Spatter) in laser powder bed fusion images. The method starts from 34 manually labeled FE-SEM images, trains a supervised CNN, and then iteratively pseudo-labels unlabeled image crops using a 50% confidence threshold over four rounds. Transfer learning from a pre-trained CNN is used to extract features. The authors report a final accuracy of 98.91% and loss of 0.0188 on an 'unlabeled' test set of 829 images, and compare against several supervised, unsupervised, and semi-supervised baselines. The central claim is that TransMatch effectively leverages scarce labeled data and large unlabeled data to achieve state-of-the-art defect detection.

Significance. If the empirical results were reliable, the paper could be of practical interest to the additive manufacturing quality-inspection community by demonstrating that standard pseudo-labeling plus transfer learning works well for FE-SEM defect classification. The manuscript also collects and preprocesses a real LPBF defect dataset, which is a useful asset. However, the significance is severely limited by three issues: (i) the methodological novelty is thin, as TransMatch reduces to well-known pseudo-labeling with a confidence threshold combined with standard transfer learning; (ii) the reported numbers are inconsistent across sections and lack statistical rigor; and (iii) the evaluation protocol does not establish that the test set is independent of the pseudo-labeled training data, so the 98.91% accuracy figure may be inflated by data leakage. There is no code, no data release, and no error bars, so the results are not independently reproducible. The comparison with prior work in Table 3 is also not apples-to-apples because defect types, imaging modalities, and evaluation splits differ.

major comments (5)
  1. [Abstract, §2.1, §2.2.1] The dataset counts are internally inconsistent. The abstract states 8,284 images; §2.2.1 states 13,484 training and 1,500 test images (total 14,984); §2.1 states 6,742/7,455 training and 750/829 test images for supervised/unsupervised paths. The relationship between these numbers, the 232 original FE-SEM images, and the 14,984 'annotations' extracted from 34 labeled images is never reconciled. This makes it impossible to know the actual data composition and undermines the reliability of all reported accuracies.
  2. [§2.1, §2.2.2, §3] The test set's independence from the pseudo-labeled training set is not established. The paper never explains how the 232 parent images were divided into labeled, unlabeled, and test crops. Because pseudo-labeling in §2.2.2 adds confidently predicted unlabeled images to the training set across four rounds, the 829-image test set reported in §3 could contain crops from the same parent images that were used as pseudo-labeled training crops. If so, the reported 98.91% accuracy reflects memorization rather than generalization. This is the weakest link in the empirical argument and must be resolved by an explicit, verifiable data-split description.
  3. [§3, Tables 1 and 2] All results are single numbers with no error bars, no repeated runs, and no variance estimates. The accuracy 0.9891 appears once, while elsewhere the text and tables report '99%' accuracy. Table 2 shows macro/weighted averages of 0.99 for 829 support, but the per-class numbers (e.g., Crack recall 0.94) imply a weighted accuracy that is not visibly consistent with 0.9891. The loss value 0.0188 is reported without a curve or definition of the split it was computed on. Statistical rigor is essential for a claim of state-of-the-art performance.
  4. [§2.2.3, §2.2.4, §3] The novelty claim is overstated. TransMatch is described as a framework that combines transfer learning with pseudo-labeling, but Algorithmically it reduces to (1) training a CNN on a small labeled set, (2) adding high-confidence predictions from an unlabeled pool to the training set, and (3) fine-tuning a pre-trained network. This is standard pseudo-labeling (Lee, 2013, cited as [38]) with a 50% threshold. The 'few-shot' and 'transMatch' language does not introduce a new mechanism or theoretical insight. The paper should either clearly position the contribution as an application study of existing methods or provide a novel algorithmic component, with an ablation to justify its necessity.
  5. [§3, Table 3] The comparison table lists prior works with widely varying defect types, imaging modalities, and evaluation protocols. Several rows use different definitions of 'accuracy' (e.g., porosity detection vs. defect classification) and some cite sources with incompatible experimental setups. The conclusion that TransMatch 'outperforms' these methods is not supported because no baseline is retrained on the same data. A fair comparison would require applying the same preprocessing and evaluation protocol to a common dataset, or at least reporting the conditions under which the comparison is meaningful.
minor comments (6)
  1. [Throughout] There are several broken cross-references in the text, including 'Error! Reference source not found.' after '99% accuracy' and after 'loss of 0.0188'. These should be fixed before resubmission.
  2. [§2.2.2] The Python snippet is mentioned but not shown in the manuscript; either include the code or remove the reference to it.
  3. [§2.1, §2.2.1] The distinction between 'supervised' and 'unsupervised' data is confusing: in §2.1 the unsupervised split has 7,455 train / 829 test while §2.2.1 says 13,484 train / 1,500 test. Clarify which of these corresponds to the pseudo-labeling rounds and which to the final SSFSL evaluation.
  4. [§3] The confusion matrix (Figure 10) is referenced but not described in detail. Add the actual matrix values or a clearer discussion of the misclassifications, especially the claim that cracks are confused with pinholes/spatter.
  5. [Conclusion] The conclusion reports per-class accuracies (93.67% for Crack, etc.) that do not match the precision/recall values in Table 2; for example, Crack precision is 1.00 and recall 0.94, giving an F1 of 0.97, not 93.67%. Reconcile these numbers.
  6. [References] The reference list contains several incomplete entries (e.g., [48] lacks full author list, [59] is a thesis without a published venue). Ensure all references are complete and properly formatted.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: TransMatch is an empirical combination of cited pseudo-labeling and transfer learning; the reported test accuracy is an independent measurement, though data-split details are underreported.

full rationale

The paper's claimed derivation chain is an empirical pipeline: supervised CNN training on extracted defect crops, four rounds of pseudo-labeling with a 50% confidence threshold, and transfer learning from a pretrained CNN. Each ingredient is explicitly attributed to prior work (Lee [38] for pseudo-labeling; Qi, Qiao, Gidaris, and Chen [45–48] for transfer-based few-shot classification), and no equation is given that would make a predicted quantity equal to a fitted quantity by construction. The central empirical claim — 98.91% accuracy on 829 images — is reported for a separately listed test subset in Section 2.1 and Table 2, not for the pseudo-labeling training pool, so it does not reduce to the training objective. The only author self-citations ([34], [35]) support general CNN-based AM monitoring and are not load-bearing for the reported result. The main weakness is that the paper does not explain how the 232 parent FE-SEM images were divided into labeled, unlabeled, and test crops, leaving a possible data-leakage risk; however, that is a validity concern, not circularity of derivation. The 'TransMatch' label is a thin re-branding of pseudo-labeling plus transfer learning, but naming a combination of cited methods does not make the empirical measurement circular.

Assumptions & free parameters 4 free parameters · 5 assumptions · 1 invented entities

The central accuracy claim depends on several unstated choices: the confidence threshold and number of rounds are free parameters, and the validity of transfer learning and pseudo-labels rests on domain assumptions about the data distribution and label quality. The test set independence assumption is especially fragile given the unclear partition of the original 232 images.

free parameters (4)
  • pseudo-label confidence threshold = 0.5 (50%)
    Governing inclusion of unlabeled images; chosen as a trade-off between training set size and pseudo-label quality (Sec 2.2.2).
  • number of pseudo-labeling rounds = 4
    Stopping condition; arbitrary, no ablation (Sec 2.2.2).
  • CNN architecture and training hyperparameters = not specified
    Layer counts, learning rate, batch size are not reported, so the model definition is a free choice by the authors (Sec 2.2.1).
  • preprocessing parameters = not specified
    GaussianBlur kernel size, adaptive threshold block size, and Canny thresholds are not given, affecting what edges are detected (Sec 2.1).
assumptions (5)
  • domain assumption Unlabeled FE-SEM images share the same distribution as the labeled images.
    Pseudo-labeling and transfer learning require the unlabeled data to be drawn from the same defect distribution as the labeled examples; stated implicitly in Sec 2.2.2.
  • domain assumption The four defect classes (Crack, Pinhole, Hole, Spatter) are mutually exclusive and the human annotations are correct.
    The dataset labels from LabelImg are taken as ground truth without inter-rater validation (Sec 2.1).
  • domain assumption ImageNet-pretrained features transfer to grayscale FE-SEM defect images.
    Transfer learning from a pre-trained CNN is assumed to help despite the domain shift from natural images to electron microscopy (Sec 2.2.4).
  • domain assumption Pseudo-labels with confidence above 50% are correct enough to improve the model.
    The iterative pseudo-labeling loop assumes that confident predictions are reliable; the paper does not verify the accuracy of the pseudo-labels (Sec 2.2.2).
  • domain assumption The test set is independent of the pseudo-labeled training set.
    Accuracy is only meaningful if no data leakage occurs, but the paper does not describe how the test patches were separated from the same parent images (Sec 2.1).
invented entities (1)
  • TransMatch framework
    purpose: Named as a novel semi-supervised few-shot learning algorithm that combines transfer learning and pseudo-labeling.
    The framework is not formally defined or released; the steps described reduce to standard pseudo-labeling and transfer learning, so it functions as a brand name rather than a new entity.

how reviews work

0 comments
Cite this review

Pith. "Pith review of TransMatch: A Transfer-Learning Framework for Defect Detection in Laser Powder Bed Fusion Additive Manufacturing." pith.science (2026). https://pith.science/paper/AWOMU4HH

@misc{pith2026250901754,
  author       = {Pith},
  title        = {Pith review of: TransMatch: A Transfer-Learning Framework for Defect Detection in Laser Powder Bed Fusion Additive Manufacturing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AWOMU4HH}},
  note         = {Machine review of arXiv:2509.01754}
}
read the original abstract

Surface defects in Laser Powder Bed Fusion (LPBF) pose significant risks to the structural integrity of additively manufactured components. This paper introduces TransMatch, a novel framework that merges transfer learning and semi-supervised few-shot learning to address the scarcity of labeled AM defect data. By effectively leveraging both labeled and unlabeled novel-class images, TransMatch circumvents the limitations of previous meta-learning approaches. Experimental evaluations on a Surface Defects dataset of 8,284 images demonstrate the efficacy of TransMatch, achieving 98.91% accuracy with minimal loss, alongside high precision, recall, and F1-scores for multiple defect classes. These findings underscore its robustness in accurately identifying diverse defects, such as cracks, pinholes, holes, and spatter. TransMatch thus represents a significant leap forward in additive manufacturing defect detection, offering a practical and scalable solution for quality assurance and reliability across a wide range of industrial applications.

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. LabelImg: CNN-Based Surface Defect Detection

    cond-mat.mes-hall 2025-09 reject novelty 4.0 of 10

    A CNN trained on a new 14,982-image LPBF dataset reportedly classifies four defect types with ~99% accuracy, but detection/segmentation claims are not actually evaluated.

Reference graph

Works this paper leans on

65 extracted references · 55 canonical work pages · cited by 1 Pith paper

  1. [38]

    Commun ACM 60:

    Krizhevsky A, Sutskever I, Hinton GE (2017) ImageNet Classification with Deep Convolutional Neural Networks. Commun ACM 60:. https://doi.org/10.1145/3065386

    doi:10.1145/3065386 2017

  2. [1]

    Hole", Figure 2(c), when the pore size exceeds 5, and conversely, it is referred to as a

    Introduction Laser Powder Bed Fusion (LPBF), an advanced additive manufacturing (AM) technique, offers an innovative method for constructing intricate metallic components directly from computer -aided design (CAD) files. This advanced process involves the meticulous melting of alloy powder in a layer-by-layer fashion, as referenced in several studies [1–3...

    work page

  3. [2]

    Specifically, the process of Selective Laser Melting (SLM) involves the complete melting of powders, as illustrated in Figure 3 (a)

    Research Methodology In the research methodology, the selective utilization of powder melting techniques plays a pivotal role. Specifically, the process of Selective Laser Melting (SLM) involves the complete melting of powders, as illustrated in Figure 3 (a). This method, implemented in Laser Powder Bed Fusion (LPBF), is characterized by the full liquefac...

    work page

  4. [3]

    The methodology initiated with the manual labeling of 232 high-resolution LPBF images, distinguishing between Crack, Pinhole, Hole, and Spatter, using the LabelImg framework

    Results and Discussion Semi-supervised learning, a powerful amalgamation of labeled and unlabeled data, provides an innovative solution to the challenges of handling limited labeled datasets. The methodology initiated with the manual labeling of 232 high-resolution LPBF images, distinguishing between Crack, Pinhole, Hole, and Spatter, using the LabelImg f...

    work page

  5. [4]

    By combining transfer learning with SSFSL, the model effectively exploits both labeled and unlabeled data to classify four main defect types—Crack, Pinhole, Hole, and Spatter

    Conclusion This paper has introduced a novel Semi-Supervised Few-Shot Learning (SSFSL) framework for surface defect detection in LPBF-built Ti alloys, leveraging a CNN pre-trained on 34 labeled images and a multi-iteration pseudo- labeling strategy. By combining transfer learning with SSFSL, the model effectively exploits both labeled and unlabeled data t...

    work page

  6. [5]

    Addit Manuf 27:42–53

    Okaro IA, Jayasinghe S, Sutcliffe C, et al (2019) Automatic fault detection for laser powder-bed fusion using semi-supervised machine learning. Addit Manuf 27:42–53. https://doi.org/10.1016/J.ADDMA.2019.01.006

    work page doi:10.1016/j.addma.2019.01.006 2019

  7. [7]

    Addit Manuf 40:101927

    Constantin L, Kraiem N, Wu Z, et al (2021) Manufacturing of complex diamond-based composite structures via laser powder- bed fusion. Addit Manuf 40:101927. https://doi.org/10.1016/J.ADDMA.2021.101927

    work page arXiv 2021

  8. [8]

    ASME International Mechanical Engineering Congress and Exposition, Proceedings (IMECE) 2B -2020:

    Song J, Moon YB (2021) Infill Defective Detection System Augmented by Semi -Supervised Learning. ASME International Mechanical Engineering Congress and Exposition, Proceedings (IMECE) 2B -2020:. https://doi.org/10.1115/IMECE2020- 23249

    work page doi:10.1115/imece2020- 2021

Show all 65 references
  1. [9]

    International Journal of Additive and Subtractive Materials Manufacturing 1:172

    Taheri H, Shoaib MRBM, Koester LW, et al (2017) Powder -based additive manufacturing - a review of types of defects, generation mechanisms, detection, property evaluation and metrology. International Journal of Additive and Subtractive Materials Manufacturing 1:172. https://do...

    2017

  2. [10]

    Journal of Materials Engineering and Performance 2021 30:7 30:4808–4818

    Brennan MC, Keist JS, Palmer TA, et al (2021) Defects in Metal Additive Manufacturing Processes. Journal of Materials Engineering and Performance 2021 30:7 30:4808–4818. https://doi.org/10.1007/S11665-021-05919-6

    2021 doi

  3. [11]

    Production Engineering 4:35–45

    Zaeh MF, Branner G (2010) Investigations on residual stresses and deformations in selective laser melting. Production Engineering 4:35–45. https://doi.org/10.1007/S11740-009-0192-Y/FIGURES/19

    2010 doi

  4. [12]

    Rapid Prototyp J 13:196–203

    Vandenbroucke B, Kruth JP (2007) Selective laser melting of biocompatible metals for rapid manufacturing of medical parts. Rapid Prototyp J 13:196–203. https://doi.org/10.1108/13552540710776142/FULL/XML

    2007 doi

  5. [13]

    Chinese Journal of Mechanical Engineering: Additive Manufacturing Frontiers 1:100037

    Yang G, Xie Y , Zhao S, et al (2022) Quality Control: Internal Defects Formation Mechanism of Selective Laser Melting Based on Laser -powder-melt Pool Interaction: A Review. Chinese Journal of Mechanical Engineering: Additive Manufacturing Frontiers 1:100037. https://doi.org/1...

    2022

  6. [14]

    J Manuf Process 75:1003 –1011

    Wang H, Gould B, Haddad M, et al (2022) In situ high -speed synchrotron X -ray imaging of laser -based directed energy deposition of the alloying process with dissimilar powders. J Manuf Process 75:1003 –1011. https://doi.org/10.1016/J.JMAPRO.2022.01.055

    2022 doi

  7. [15]

    J Appl Phys 93:10089–10096

    Zhao H, DebRoy T (2003) Macroporosity free aluminum alloy weldments through numerical simulation of keyhole mode laser welding. J Appl Phys 93:10089–10096. https://doi.org/10.1063/1.1573732

    2003 doi

  8. [16]

    International Journal of Advanced Manufacturing Technology 120:5117 –5129

    Ilani MA, Khoshnevisan M (2022) An evaluation of the surface integrity and corrosion behavior of Ti -6Al-4 V processed thermodynamically by PM -EDM criteria. International Journal of Advanced Manufacturing Technology 120:5117 –5129. https://doi.org/10.1007/S00170-022-09093-4/F...

    2022 doi

  9. [17]

    International Journal of Advanced Manufacturing Technology 116:1763–1782

    Ilani MA, Khoshnevisan M (2021) Study of surfactant effects on intermolecular forces (IMF) in powder -mixed electrical discharge machining (EDM) of Ti -6Al-4V . International Journal of Advanced Manufacturing Technology 116:1763–1782. https://doi.org/10.1007/S00170-021-07569-3...

    2021 doi

  10. [18]

    Nature Communications 2022 13:1 13:1–12

    Poudel A, Yasin MS, Ye J, et al (2022) Feature-based volumetric defect classification in metal additive manufacturing. Nature Communications 2022 13:1 13:1–12. https://doi.org/10.1038/s41467-022-34122-x

    2022 doi

  11. [19]

    J Manuf Process 36:442–449

    Mukherjee T, DebRoy T (2018) Mitigation of lack of fusion defects in powder bed fusion additive manufacturing. J Manuf Process 36:442–449. https://doi.org/10.1016/J.JMAPRO.2018.10.028

    2018 doi

  12. [20]

    Nature Communications 2022 13:1 13:1–11

    Huang Y , Fleming TG, Clark SJ, et al (2022) Keyhole fluctuation and pore formation mechanisms during laser powder bed fusion additive manufacturing. Nature Communications 2022 13:1 13:1–11. https://doi.org/10.1038/s41467-022-28694-x

    2022 doi

  13. [21]

    Materials 2018, V ol 11, Page 106 11:106

    Marchese G, Basile G, Bassini E, et al (2018) Study of the Microstructure and Cracking Mechanisms of Hastelloy X Produced by Laser Powder Bed Fusion. Materials 2018, V ol 11, Page 106 11:106. https://doi.org/10.3390/MA11010106

    2018 doi

  14. [22]

    Materials Science and Engineering: A 732:228 –239

    Han Q, Mertens R, Montero-Sistiaga ML, et al (2018) Laser powder bed fusion of Hastelloy X: Effects of hot isostatic pressing and the hot cracking mechanism. Materials Science and Engineering: A 732:228 –239. https://doi.org/10.1016/J.MSEA.2018.07.008

    2018 doi

  15. [23]

    Mater Des 104:404 –413

    Reichardt A, Dillon RP, Borgonia JP, et al (2016) Development and characterization of Ti -6Al-4V to 304L stainless steel gradient components fabricated with laser deposition additive manufacturing. Mater Des 104:404 –413. https://doi.org/10.1016/J.MA TDES.2016.05.016

    2016 doi

  16. [24]

    International Journal of Advanced Manufacturing Technology 90:561–574

    Gaja H, Liou F (2017) Defects monitoring of laser metal deposition using acoustic emission sensor. International Journal of Advanced Manufacturing Technology 90:561–574. https://doi.org/10.1007/S00170-016-9366-X/METRICS

    2017 doi

  17. [25]

    Welding Metallurgy and Weldability 9781118230701:1 –400

    Lippold JC (2014) Welding Metallurgy and Weldability. Welding Metallurgy and Weldability 9781118230701:1 –400. https://doi.org/10.1002/9781118960332

    2014 doi

  18. [26]

    Metallurgical Transactions B 1976 7:4 7:599–606

    Fredriksson H, Svensson I (1976) On the mechanism of pore formation in metals. Metallurgical Transactions B 1976 7:4 7:599–606. https://doi.org/10.1007/BF02698593

    1976 doi

  19. [27]

    J Mater Process Technol 214:2915–2925

    King WE, Barth HD, Castillo VM, et al (2014) Observation of keyhole-mode laser melting in laser powder-bed fusion additive manufacturing. J Mater Process Technol 214:2915–2925. https://doi.org/10.1016/J.JMATPROTEC.2014.06.005

    2014 doi

  20. [28]

    J Laser Appl 10:247 –

    Matsunawa A, Kim J-D, Seto N, et al (1998) Dynamics of keyhole and molten pool in laser welding. J Laser Appl 10:247 –

    1998

  21. [29]

    J Intell Manuf 1–16

    Lui CF, Maged A, Xie M (2023) A novel image feature based self-supervised learning model for effective quality inspection in additive manufacturing. J Intell Manuf 1–16. https://doi.org/10.1007/S10845-023-02232-Y/FIGURES/9

    2023 doi

  22. [30]

    J Laser Appl 13:187–192

    Katayama S, Kobayashi Y , Mizutani M, Matsunawa A (2001) Effect of vacuum on penetration and defects in laser welding. J Laser Appl 13:187–192. https://doi.org/10.2351/1.1404413

    2001 doi

  23. [31]

    Addit Manuf 36:

    Young ZA, Guo Q, Parab ND, et al (2020) Types of spatter and their features and formation mechanisms in laser powder bed fusion additive manufacturing process. Addit Manuf 36:. https://doi.org/10.1016/J.ADDMA.2020.101438

    2020

  24. [32]

    International Journal of Pavement Research and Technology 16:943–967

    Nguyen SD, Tran TS, Tran VP, et al (2023) Deep Learning -Based Crack Detection: A Survey. International Journal of Pavement Research and Technology 16:943–967. https://doi.org/10.1007/S42947-022-00172-Z/TABLES/6

    2023 doi

  25. [33]

    IEEE Trans Industr Inform 18:6743–6755

    Yang B, Liu Z, Duan G, Tan J (2022) Mask2Defect: A Prior Knowledge-Based Data Augmentation Method for Metal Surface Defect Inspection. IEEE Trans Industr Inform 18:6743–6755. https://doi.org/10.1109/TII.2021.3126098

    2022

  26. [34]

    https://doi.org/101117/122513816 10969:50–57

    Razaviarab N, Sharifi S, Banadaki YM (2019) Smart additive manufacturing empowered by a closed -loop machine learning algorithm. https://doi.org/101117/122513816 10969:50–57. https://doi.org/10.1117/12.2513816

    2019 doi

  27. [35]

    Comput Ind 140:103662

    Wang R, Cheung CF, Wang C, Cheng MN (2022) Deep learning characterization of surface defects in the selective laser melting process. Comput Ind 140:103662. https://doi.org/10.1016/J.COMPIND.2022.103662

    2022

  28. [36]

    Rapid Prototyp J 20:77–86

    Barua S, Liou F, Newkirk J, Sparks T (2014) Vision-based defect detection in laser metal deposition process. Rapid Prototyp J 20:77–86. https://doi.org/10.1108/RPJ-04-2012-0036

    2014 doi

  29. [37]

    Nature 2015 521:7553 521:436–444

    Lecun Y , Bengio Y , Hinton G (2015) Deep learning. Nature 2015 521:7553 521:436–444. https://doi.org/10.1038/nature14539

    2015 doi

  30. [39]

    Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 11209 LNCS:311–327

    Shi W, Gong Y , Ding C, et al (2018) Transductive Semi-Supervised Deep Learning Using Min-Max Features. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 11209 LNCS:311–327. https://doi.org/10.1...

    2018 doi

  31. [40]

    Recent Progress in Materials 2022, V ol 4, 005 4:1 –19

    Yaser B, Nariman R, Hadi F, et al (2022) Automated Quality and Process Control for Additive Manufacturing using Deep Convolutional Neural Networks. Recent Progress in Materials 2022, V ol 4, 005 4:1 –19. https://doi.org/10.21926/RPM.2201005

    2022 doi

  32. [41]

    The International Journal of Advanced Manufacturing Technology 2021 113:1 113:35 –58

    Zheng X, Zheng S, Kong Y , Chen J (2021) Recent advances in surface defect inspection of industrial products using deep learning techniques. The International Journal of Advanced Manufacturing Technology 2021 113:1 113:35 –58. https://doi.org/10.1007/S00170-021-06592-8

    2021 doi

  33. [42]

    J Intell Manuf 34:3091–3108

    Manivannan S (2023) Automatic quality inspection in additive manufacturing using semi -supervised deep learning. J Intell Manuf 34:3091–3108. https://doi.org/10.1007/S10845-022-02000-4/FIGURES/11

    2023 doi

  34. [43]

    Lee D-H (2013) Pseudo-Label : The Simple and Efficient Semi-Supervised Learning Method for Deep Neural Networks

    2013

  35. [44]

    International Congress on Ultra Modern Telecommunications and Control Systems and Workshops 2021-October:60–65

    Kumar A, Joshi RC, Dutta MK, et al (2021) Fruit -CNN: An Efficient Deep learning -based Fruit Classification and Quality Assessment for Precision Agriculture. International Congress on Ultra Modern Telecommunications and Control Systems and Workshops 2021-October:60–65. https:...

    2021

  36. [45]

    Virtual Phys Prototyp 18:

    Nguyen NV , Hum AJW, Do T, Tran T (2023) Semi-supervised machine learning of optical in-situ monitoring data for anomaly detection in laser powder bed fusion. Virtual Phys Prototyp 18:. https://doi.org/10.1080/17452759.2022.2129396

    2023

  37. [46]

    J Intell Manuf 33:457–471

    Larsen S, Hooper PA (2022) Deep semi-supervised learning of dynamics for anomaly detection in laser powder bed fusion. J Intell Manuf 33:457–471. https://doi.org/10.1007/S10845-021-01842-8/FIGURES/10

    2022 doi

  38. [47]

    Multimed Tools Appl 78:3767 –

    Ma Y , Liu Y , Xie Q, Li L (2019) CNN-feature based automatic image annotation method. Multimed Tools Appl 78:3767 –

    2019

  39. [48]

    Chen W-Y , Liu Y-C, Kira Z, et al A CLOSER LOOK AT FEW-SHOT CLASSIFICATION

  40. [49]

    Comput Ind 148:103911

    Jha SB, Babiceanu RF (2023) Deep CNN-based visual defect detection: Survey of current literature. Comput Ind 148:103911. https://doi.org/10.1016/J.COMPIND.2023.103911

    2023

  41. [50]

    Journal of Manufacturing Science and Engineering, Transactions of the ASME 140:

    Imani F, Gaikwad A, Montazeri M, et al (2018) Process mapping and in -process monitoring of porosity in laser powder bed fusion using layerwise optical imaging. Journal of Manufacturing Science and Engineering, Transactions of the ASME 140:. https://doi.org/10.1115/1.4040615/366215

    2018 doi

  42. [51]

    Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition 5822–5830

    Qi H, Brown M, Lowe DG (2017) Low -Shot Learning with Imprinted Weights. Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition 5822–5830. https://doi.org/10.1109/CVPR.2018.00610

    2017

  43. [52]

    Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition 7229 –7238

    Qiao S, Liu C, Shen W, Yuille A (2017) Few-Shot Image Recognition by Predicting Parameters from Activations. Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition 7229 –7238. https://doi.org/10.1109/CVPR.2018.00755

    2017

  44. [53]

    Paris-Est U, Komodakis N Dynamic Few-Shot Visual Learning without Forgetting Spyros Gidaris

  45. [54]

    Smart Sustain Manuf Syst 3:98–121

    Gaikwad A, Imani F, Yang H, et al (2019) In Situ Monitoring of Thin-Wall Build Quality in Laser Powder Bed Fusion Using Deep Learning. Smart Sustain Manuf Syst 3:98–121. https://doi.org/10.1520/SSMS20190027

    2019 doi

  46. [55]

    Petrich J, Gobert C, Phoha S, et al (2017) Machine Learning for Defect Detection for PBFAM Using High Resolution Layerwise Imaging Coupled with Post-Build CT Scans

    2017

  47. [56]

    Progress in Additive Manufacturing 5:277 –285

    Baumgartl H, Tomas J, Buettner R, Merkel M (2020) A deep learning -based model for defect detection in laser -powder bed fusion using in -situ thermographic monitoring. Progress in Additive Manufacturing 5:277 –285. https://doi.org/10.1007/S40964-019-00108-3/FIGURES/4

    2020 doi

  48. [57]

    Applied Sciences 2020, V ol 10, Page 545 10:545

    Cui W, Zhang Y , Zhang X, et al (2020) Metal Additive Manufacturing Parts Inspection Using Convolutional Neural Network. Applied Sciences 2020, V ol 10, Page 545 10:545. https://doi.org/10.3390/APP10020545

    2020 doi

  49. [58]

    Addit Manuf 28:497–505

    Zhang B, Liu S, Shin YC (2019) In-Process monitoring of porosity during laser additive manufacturing process. Addit Manuf 28:497–505. https://doi.org/10.1016/J.ADDMA.2019.05.030

    2019 doi

  50. [59]

    CIRP Ann Manuf Technol 69:205 –208

    Guo W “Grace,” Tian Q, Guo S, Guo Y (2020) A physics-driven deep learning model for process-porosity causal relationship and porosity prediction with interpretability in laser metal deposition. CIRP Ann Manuf Technol 69:205 –208. https://doi.org/10.1016/J.CIRP.2020.04.049

    2020 doi

  51. [60]

    Virtual Phys Prototyp 16:481–497

    Pandiyan V , Drissi-Daoudi R, Shevchik S, et al (2021) Semi-supervised Monitoring of Laser powder bed fusion process based on acoustic emissions. Virtual Phys Prototyp 16:481–497. https://doi.org/10.1080/17452759.2021.1966166

    2021

  52. [61]

    Journal of Manufacturing Science and Engineering, Transactions of the ASME 141:

    Imani F, Chen R, Diewald E, et al (2019) Deep learning of variant geometry in layerwise imaging profiles for additive manufacturing quality control. Journal of Manufacturing Science and Engineering, Transactions of the ASME 141:. https://doi.org/10.1115/1.4044420/956249

    2019 doi

  53. [63]

    https://doi.org/10.26153/TSW/17561

    Zhao X, Imandoust A, Khanzadeh M, et al (2021) Automated Anomaly Detection of Laser -Based Additive Manufacturing Using Melt Pool Sparse Representation and Unsupervised Learning. https://doi.org/10.26153/TSW/17561

    2021 doi

  54. [64]

    Extreme Mech Lett 41:101029

    Guo K, Buehler MJ (2020) A semi-supervised approach to architected materials design using graph neural networks. Extreme Mech Lett 41:101029. https://doi.org/10.1016/J.EML.2020.101029

    2020

  55. [65]

    Göransson A (2022) Defect classification in LPBF images using semi-supervised learning

    2022

  56. [254]

    https://doi.org/10.2351/1.521858

    doi

  57. [3780]

    https://doi.org/10.1007/S11042-018-6038-X

    doi

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.