Pith. sign in

REVIEW 3 major objections 6 minor 36 references

A Generative Approach for Improving Multi-Label Defect Classification in Photovoltaic Modules

T0 review · 3 major / 6 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read This paper shows that Generative Defect Isolation (GDI) — erasing non-target defects via LaMa inpainting — improves multi-label photovoltaic defect classification, raising rare-class F1 by up to 63.6% and cutting co-occurring errors by 26%.

desk verdict GDI is a genuinely useful data-centric augmentation with aggregate gains that survive a fixed-threshold check, but the headline rare-class and co-occurrence numbers come from test-set threshold tuning and are less solid than the abstract suggests. read the letter →

arxiv 2608.12725 v1 pith:AW2YKMNH submitted 2026-08-13 cs.CV

Abdul Mueez , Yogesh S. Rawat , Shruti Vyas This is my paper · ORCID
classification cs.CV
keywords PhotovoltaicsElectroluminescenceimagingMulti-labeldefectclassificationSyntheticdataInpaintingGenerativeIsolationaugmentation
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

The paper asks whether the ambiguity of multi-label defect classification in electroluminescence (EL) images of solar cells can be reduced by changing the training data rather than the model. It proposes Generative Defect Isolation (GDI): for images with several co-occurring defects, the ground-truth segmentation masks are used to erase all but one defect with the LaMa inpainting model, producing clean single-defect examples and one defect-free example from each source image. Across ViT-S, ViT-L, and EfficientNetV2-L, training on the original plus GDI-augmented images improves zero-one accuracy and macro F1 in nearly every data-scarce setting, with the largest gains on rare classes (up to +63.6% F1) and a 26% reduction in co-occurring classification errors. The authors position GDI as a way to repurpose expensive segmentation annotations into faster, production-ready multi-label classifiers at no inference-time cost.

What carries the argument

The central object is the GDI pipeline built on LaMa (Large Mask Inpainting), a neural inpainting network whose bottleneck uses Fast Fourier Convolution residual blocks: one branch processes local texture with ordinary convolutions while the other applies a real FFT, a $1\times 1$ convolution, and an inverse FFT to obtain an image-wide receptive field that can reproduce the periodic grid lines of PV cells. GDI selects a target defect only if its annotated area exceeds 10% of the image, dilates the mask of all other defects with a $15\times 15$ kernel, and asks LaMa to fill the masked region, then repeats for each qualifying defect and finally generates a no-defect sample by inpainting all defects at once. The augmented single-defect images are one-hot labeled and added to the original multi-defect training set, so the classifier sees both clean single-defect mappings and real co-occurrence patterns.

What would settle it

Filter out every GDI-generated no-defect image whose inpainted region fails to reconstruct grid lines (the Figure 8-type case) and retrain the same classifiers; if the macro F1 and zero-one accuracy gains over baseline persist, the benefit comes from defect isolation itself, whereas if they collapse to noise, the claimed gains were carried by inpainting fidelity rather than by removing co-occurrence ambiguity.

Watch

Extended reading notes

Core claim

GDI is the claim that annotating which pixels belong to each defect lets a generative inpainting model subtract defects from real EL images, yielding training examples in which one defect appears alone or no defect appears at all, and that training on this hybrid set disentangles visual features better than training on the original multi-defect images alone. In the paper's strongest results, GDI raises macro F1 from 0.7672 to 0.7744 and zero-one accuracy from 0.5943 to 0.6046 for EfficientNetV2-L on the full dataset, and at the 20% data split it lifts rare-class F1 by 63.6% (Contact_BeltMarks) and 20.1% (Crack_Isolated). The error-co-occurrence matrices show the total number of paired class errors falling from 1,774 to 1,312, a 26% drop, with the Unknown-class confusions decreasing most sharply; multi-seed experiments at a fixed threshold of $\tau=0.5$ confirm the gains are not an artifact of per-class threshold tuning.

Load-bearing premise

GDI's gains depend on LaMa, pretrained on natural images, producing inpainted EL-cell regions that are visually and structurally faithful enough that the clean single-defect and no-defect examples transfer to real test cells — the paper itself acknowledges a failure mode when combined defect masks hide all grid lines.

Editorial extensions

If this is right

  • Rare defect classes become learnable from just a handful of clean, inpainted single-defect examples: Contact_BeltMarks, with only 6 original labels in the 20% split, gains +63.6% relative F1 after 3 GDI samples are added.
  • GDI's benefit is largest when training data is scarce — e.g., +125.1% zero-one accuracy for ViT-S at 10% data — and shrinks but stays positive at full data, making the method most valuable where annotation budgets are tight.
  • Because the Unknown class gains +16.5% F1 without any new Unknown samples, GDI indirectly reduces a model's overuse of ambiguous labels by drawing clearer boundaries for the known defect classes.
  • GDI is a one-time offline preprocessing step with no inference-time cost, so the expensive pixel-level segmentation knowledge is distilled offline into a fast classifier that can screen and bin cells on a production line.
  • GDI outperforms Copy-Paste augmentation on all three architectures, and the paper attributes this to the physical implausibility of pasting defects whose locations are constrained by busbar stress and contact interfaces.

Reading between the lines

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

  • Going beyond the paper's experiments, GDI should transfer to other defect domains where segmentation masks already exist and the background has strong structural regularity — pipe welds, textiles, or metal surfaces — because the core mechanism, removing confounded co-occurring labels from training images, is not specific to photovoltaics.
  • The fixed-threshold multi-seed results (11 of 12 accuracy improvements, 10 of 12 F1 improvements) suggest a practical deployment rule the paper does not state: apply GDI before searching over architectures, since the smallest models receive the largest relative lift.
  • A testable extension is to filter out the acknowledged grid-line failure cases (Figure 8) from the generated No_Defect set and retrain; if the gains persist, defect isolation itself is the driver, whereas if they shrink, inpainting fidelity is the true source of the improvement.
  • The authors name a domain-specific inpainting model pretrained on PV cell topology as future work; if such a model eliminates the grid-line failure mode, the likely downstream effect is a further drop in Unknown-class confusion rather than a large change in already-well-classified common defects.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. This paper introduces Generative Defect Isolation (GDI), an annotation-guided data augmentation technique for multi-label defect classification in electroluminescence (EL) images of photovoltaic cells. GDI uses the LaMa inpainting model to remove selected defects from multi-defect training images, generating single-defect and no-defect training samples from existing segmentation annotations. The authors evaluate GDI on the UCF-EL-Defect dataset with ViT-S, ViT-L, and EfficientNetV2-L classifiers, reporting that GDI improves zero-one accuracy and macro F1 across several training-data fractions, with the largest relative gains in low-data regimes. They further report a 63.6% F1 gain for the rare Contact_BeltMarks class and a 26% reduction in co-occurring error pairs, and they compare against copy-paste augmentation and re-implemented baselines from prior work. The main experiments select per-class decision thresholds on the test set, a fact disclosed in Section 4.2.2, and Section 4.9 provides a fixed-threshold, multi-seed robustness check for aggregate metrics.

Significance. GDI is a practical and clearly described idea: repurposing pixel-level segmentation annotations to create cleaner training examples for a downstream classification task, with no additional inference cost. The paper is honest about the test-set threshold selection in the main protocol and includes a complementary fixed-threshold multi-seed experiment that supports the aggregate claim that GDI improves mean macro F1 and zero-one accuracy across all three architectures. The comparison with copy-paste augmentation and the inclusion of code and synthetic images are useful contributions. However, the stress-test concern is valid: the headline rare-class and co-occurrence-disambiguation numbers come from threshold-tuned single runs, and the fixed-threshold study does not cover class-wise or co-occurrence metrics. The mechanism-level claims therefore need additional support before the paper can be accepted as stated.

major comments (3)
  1. [Section 4.6, Table 3, Figure 5] The +63.6% relative F1 gain for Contact_BeltMarks is computed under the main protocol in which per-class thresholds are selected to maximize macro F1 on the test set, and this class has only 5 test samples and 3 inpainted training samples in the 20% split. Threshold selection can therefore move F1 by a large relative amount based on one or two prediction changes, and the fixed-threshold experiment in Section 4.9 reports only aggregate metrics, not class-wise F1. Please report class-wise F1 at the fixed threshold tau=0.5 across the four seeds, or with thresholds chosen on a validation split, so that the rare-class claim is separated from threshold overfitting.
  2. [Section 4.6, Figure 6] The 26% reduction in co-occurring error pairs and the specific pair-level reductions (e.g., Unknown/Contact_NearSolderPad falling from 165 to 66) are also obtained from the test-set-threshold-tuned run. Since the co-occurrence analysis is the main evidence for the claimed mechanism of resolving learning ambiguity, it should be recomputed at the fixed threshold (tau=0.5) or with validation-based thresholds for all seeds. Without that, the aggregate-only robustness check in Section 4.9 cannot support the co-occurrence-disambiguation part of the central claim.
  3. [Section 4.6, Figure 6] The definition of an 'error' in the co-occurrence matrix is ambiguous. The text states that an off-diagonal cell counts samples where the model made an error on both classes, but the subsequent example of the model 'co-predicting' Unknown on Contact_NearSolderPad images suggests false positives alone. Please specify whether the matrix counts false positives, false negatives, or both, and make the example consistent with that definition, because the 26% figure and the 60% reduction for the Unknown/Contact_NearSolderPad pair are headline quantitative claims.
minor comments (6)
  1. [Section 4.6, Figure 6] The sum of the upper triangle of the baseline matrix as printed is 1759, not 1774 as stated in the text; please verify the printed matrix or the reported sum.
  2. [Section 4.2.2 / Abstract / Section 6] The abstract and conclusion repeat the +63.6% and 26% figures without noting that they come from the test-set-threshold-tuned protocol; consider adding that caveat wherever these numbers appear.
  3. [Section 4.7, Table 4] The comparison with prior work re-implements their backbones under the authors' 12-class label space and a new 80/20 split; the 'new state-of-the-art' claim should be framed as relative to these re-implemented baselines rather than to the original published results.
  4. [Section 3.2.3] The sentence 'The network is trained using a combined objective function...' describes the pre-training of LaMa, not training performed in this paper; rephrase to avoid implying that the authors trained the inpainting model.
  5. [Section 4.3] The phrase '100% drop in accuracy' is unclear; state the actual accuracy values and describe the change as a relative change of -100%.
  6. [Section 3.2.1 / Section 3.2.2] The 10% area filter and the 15x15 dilation kernel are chosen by qualitative inspection; a sensitivity analysis would make the method's dependence on these hyperparameters explicit.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation found: GDI is an augmentation method evaluated on held-out original images, with disclosed test-set threshold tuning and a fixed-threshold robustness check.

full rationale

The paper's claimed derivation chain is empirical rather than definitional. GDI takes existing ground-truth segmentation masks from the UCF-EL-Defect dataset, inpaints selected defects with the pretrained LaMa model, and adds the resulting single-defect or no-defect images to the training set. The classifier is then trained on this augmented set and evaluated on a held-out 20% split consisting exclusively of original, unaugmented images, so the prediction target (multi-label defect presence on real test images) is never used to construct the training signal or the synthetic labels. No equation defines a predicted quantity in terms of the fitted quantity; no fitted parameter is renamed as a prediction. The per-class thresholds used in the main experiments are selected on the test set, and the paper explicitly acknowledges that this can give optimistic absolute F1 estimates. This is a statistical reporting concern, not circularity, and it is mitigated by the Section 4.9 multi-seed experiment at a fixed threshold tau = 0.5, which uses no test-set information and still shows aggregate gains for all three architectures. The class-level and co-occurrence claims rest on the threshold-tuned 20% run, where small per-class samples (e.g., Contact_BeltMarks with 5 test samples) make the magnitude of specific gains fragile; that is a robustness limitation, not a circular reduction. Self-citations to the dataset paper [8] and the semi-supervised paper [21] are normal references to public, externally available data and prior context; they do not carry a derivational premise that reduces to the paper's own conclusion. The claim of a new benchmark is a comparative empirical result on an independent test split, not a benchmark-derived prediction. Overall, no step in the paper's argument is equivalent to its input by construction.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

No new physical or mathematical entities are introduced. The central claim stands on the quality of synthetic inpainting, the accuracy of existing annotations, and the chosen experimental protocol. The free parameters listed above are the main hand-set or test-set-fitted choices that influence the reported numbers.

free parameters (3)
  • Area filter threshold = 10% of image area
    Defects whose annotated area is below 10% of the image are not isolated; this hand-set threshold determines which training samples are generated and affects the composition of the augmented set (Section 3.2.1).
  • Dilation kernel size = 15x15
    Chosen by qualitative inspection to balance defect boundary coverage and texture preservation; affects inpainting quality and the appearance of synthetic images (Section 3.2.2).
  • Per-class decision thresholds = Not reported; chosen to maximize macro F1 on the test set
    Used for all headline results in Figures 4 and 5 and Table 4; fitting thresholds on the test set can inflate absolute performance, though a fixed-threshold robustness check is provided in Section 4.9.
assumptions (3)
  • domain assumption LaMa inpainting pretrained on natural images generalizes to electroluminescence PV cell images
    The GDI pipeline relies on LaMa generating plausible PV textures and grid structures; the paper shows a failure case in Figure 8 where it does not.
  • domain assumption The segmentation annotations in UCF-EL-Defect correctly locate defect pixels
    GDI uses these masks to construct inpainting masks; any annotation error directly changes the synthetic training data (Section 3.2.2).
  • domain assumption ImageNet-pretrained weights are a useful starting point for EL images
    All models are initialized with ImageNet-1k pretrained weights, a standard but unverified assumption for this domain (Section 4.2.1).

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Generative Approach for Improving Multi-Label Defect Classification in Photovoltaic Modules." pith.science (2026). https://pith.science/paper/AW2YKMNH

@misc{pith2026260812725,
  author       = {Pith},
  title        = {Pith review of: A Generative Approach for Improving Multi-Label Defect Classification in Photovoltaic Modules},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AW2YKMNH}},
  note         = {Machine review of arXiv:2608.12725}
}
read the original abstract

This paper addresses the challenge of multi-label defect classification in electroluminescence (EL) images of photovoltaic (PV) cells. Training models on images where multiple defects co-occur creates learning ambiguity, making it difficult to disentangle visual features for specific defect types, a problem compounded by the scarcity of examples for individual classes. To tackle this, we introduce Generative Defect Isolation (GDI), utilizing the LaMa inpainting model with Fast Fourier Convolutions to remove selected defects and generate realistic, single-defect training samples. Extensive experiments on Vision Transformer (ViT-S, ViT-L) and EfficientNetV2-L architectures demonstrate that GDI significantly outperforms baselines. The performance gains are most pronounced in low-data scenarios; class-wise analysis shows substantial improvements, boosting the F1-Score for rare defect classes by up to 63.6%. Furthermore, GDI effectively resolves learning ambiguity from co-occurring defects, yielding a 26% reduction in such co-occurring classification errors. Our work establishes GDI as an effective method for maximizing the value of existing segmentation datasets and sets a new performance benchmark for multi-label classification in this domain.

Figures

Figures reproduced from arXiv: 2608.12725 by the authors.

Figure 1
Figure 1. The proposed Generative Defect Isolation (GDI) process: For images with multiple defects, the method first iteratively isolates each defect by inpainting all others. After this loop, it generates a final “no-defect” sample by creating a combined mask of all defects and inpainting them simultaneously. This creates multiple single￾defect samples and one defect-free sample from a single source image. 3.2.1. Defect Sele… view at source ↗
Figure 2
Figure 2. The proposed inpainting workflow: The original image (x) and the binary mask of other defects (Mother) are combined to create a masked input tensor (x ′ ). This tensor is processed by an inpainting network featuring an encoder-decoder architecture: three downscale blocks reduce the spatial resolution, a core of nine Fast Fourier Convolu￾tion (FFC) residual blocks processes the features and three upscale blocks recon… view at source ↗
Figure 3
Figure 3. Visual examples of the Generative Defect Isolation (GDI) pipeline. Each row corresponds to a different source image. For each, we show the original multi￾defect cell (left), the annotated defect boundaries and the high-quality samples generated by GDI. These include images with an isolated defect with an area greater than 10% of the entire image or a clean No_Defect version created by inpainting all defects. Subfigu… view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Performance comparison of the baseline and GDI models across three architectures for the six different training splits: 1%, 5%, 10%, 20%, 50%, 100%. The top image in each column shows Zero-One Accuracy and the bottom image shows Macro F1 Score. The composition of the G…
Figure 5
Figure 5. Figure 5: Class-wise F1-Scores: Comparison between the baseline and GDI-enhanced EfficientNetV2-L models on the 20% data split, showing gains for rare defects like Con￾tact_BeltMarks and Contact_Corrosion. C-BM C-Co C-FGI C-NSP K-Cl K-Is K-Re I-BS I-Dc I-HR ND Unk Class C-BM C-C…
Figure 6
Figure 6. Figure 6: Error co-occurrence matrices for the EfficientNetV2-L model on the 20% data split. Each cell (i, j) shows the number of times class i and class j were both incorrectly predicted for the same image. The GDI-augmented model (b) shows a marked reduction in error counts ac…
Figure 7
Figure 7. Figure 7: Visual samples generated via Copy-Paste augmentation. The images exhibit visible texture discontinuities and artificial boundaries where defects were super￾imposed random seeds (24, 42, 67, 76) and evaluated every run at a fixed default deci￾sion threshold of τ = 0.5. …
Figure 8
Figure 8. Figure 8: An example of a GDI failure case. When multiple defect annotations completely occlude a key structural feature, such as all the grid lines of the cell (b), the inpainting model lacks the context to reconstruct it. While it can successfully isolate a single defect - in …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 27 canonical work pages

  1. [1]

    Hijjawi, S

    U. Hijjawi, S. Lakshminarayana, T. Xu, G. P. Malfense Fierro, M. Rah- man, A review of automated solar photovoltaic defect detection systems: Approaches, challenges, andfutureorientations, SolarEnergy266(2023) 112186

    work page 2023

  2. [2]

    M. Y. Demirci, N. Beşli, A. Gümüşçü, An improved hybrid solar cell defect detection approach using generative adversarial networks and weighted classification, Expert Systems with Applications 252 (2024) 124230. 30

    work page 2024

  3. [3]

    W. Tang, Q. Yang, Z. Dai, W. Yan, Module defect detection and diagno- sis for intelligent maintenance of solar photovoltaic plants: Techniques, systems and perspectives, Energy 297 (2024) 131222

    work page 2024

  4. [4]

    Chindarkar, S

    A. Chindarkar, S. Priyadarshi, N. S. Shiradkar, A. Kottantharayil, R. Velmurugan, Deep learning based detection of cracks in electrolumi- nescence images of fielded pv modules, in: 2020 IEEE 47th Photovoltaic Specialists Conference (PVSC), IEEE, 2020, pp. 1612–1616

    work page 2020

  5. [5]

    Köntges, S

    M. Köntges, S. Kurtz, C. Packard, U. Jahn, K. Berger, K. Kato, Review of failures of photovoltaic modules, IEA-PVPS T 13 (01) (2014)

    work page 2014

  6. [6]

    Fuyuki, A

    T. Fuyuki, A. Kitiyanan, Photographic diagnosis of crystalline silicon solar cells utilizing electroluminescence, Applied physics A 96 (2009) 189–196

    work page 2009

  7. [7]

    Deitsch, V

    S. Deitsch, V. Christlein, S. Berger, C. Buerhop-Lutz, A. Maier, F. Gall- witz, C. Riess, Automatic classification of defective photovoltaic module cells in electroluminescence images, Solar Energy 185 (2019) 455–468

    work page 2019

  8. [8]

    Fioresi, D

    J. Fioresi, D. J. Colvin, R. Frota, R. Gupta, M. Li, H. P. Seigneur, S. Vyas, S. Oliveira, M. Shah, K. O. Davis, Automated defect detection and localization in photovoltaic cells using semantic segmentation of electroluminescence images, IEEE Journalof Photovoltaics 12 (1)(2022) 53–61

    work page 2022

Show all 36 references
  1. [9]

    B. Su, Z. Zhou, H. Chen, Pvel-ad: A large-scale open-world dataset for photovoltaic cell anomaly detection, IEEE Transactions on Industrial Informatics 19 (1) (2022) 404–413

    2022

  2. [10]

    M. W. Akram, G. Li, Y. Jin, X. Chen, C. Zhu, X. Zhao, A. Khaliq, M. Faheem, A. Ahmad, Cnn based automatic detection of photovoltaic cell defects in electroluminescence images, Energy 189 (2019) 116319

    2019

  3. [11]

    Suvorov, E

    R. Suvorov, E. Logacheva, A. Mashikhin, A. Remizova, A. Ashukha, A. Silvestrov, N. Kong, H. Goka, K. Park, V. Lempitsky, Resolution- robust large mask inpainting with fourier convolutions, in: Proceedings of the IEEE/CVF winter conference on applications of computer vision, 20...

    2022

  4. [12]

    Tsai, S.-C

    D.-M. Tsai, S.-C. Wu, W.-C. Li, Defect detection of solar cells in elec- troluminescence images using fourier image reconstruction, Solar energy materials and solar cells 99 (2012) 250–262

    2012

  5. [13]

    Tsai, S.-C

    D.-M. Tsai, S.-C. Wu, W.-Y. Chiu, Defect detection in solar modules usingicabasisimages, IEEETransactionsonIndustrialInformatics9(1) (2013) 122–131

    2013

  6. [14]

    Zhang, X

    J. Zhang, X. Chen, H. Wei, K. Zhang, A lightweight network for pho- tovoltaic cell defect detection in electroluminescence images based on neural architecture search and knowledge distillation, Applied Energy 355 (2024) 122184

    2024

  7. [15]

    T. Fan, T. Sun, X. Xie, H. Liu, Z. Na, Automatic micro-crack detection of polycrystalline solar cells in industrial scene, IEEE Access 10 (2022) 16269–16282

    2022

  8. [16]

    X. Zhao, C. Song, H. Zhang, X. Sun, J. Zhao, Hrnet-based automatic identification of photovoltaic module defects using electroluminescence images, Energy 267 (2023) 126605

    2023

  9. [17]

    Zhang, Y

    X. Zhang, Y. Hao, H. Shangguan, P. Zhang, A. Wang, Detection of surface defects on solar cells by fusing multi-channel convolution neural networks, Infrared Physics & Technology 108 (2020) 103334

    2020

  10. [18]

    Y. Zhao, K. Zhan, Z. Wang, W. Shen, Deep learning-based automatic detection of multitype defects in photovoltaic modules and application in real production line, Progress in Photovoltaics: Research and Appli- cations 29 (4) (2021) 471–484

    2021

  11. [19]

    Bartler, L

    A. Bartler, L. Mauch, B. Yang, M. Reuter, L. Stoicescu, Automated detection of solar cell defects with deep learning, in: 2018 26th european signal processing conference (EUSIPCO), IEEE, 2018, pp. 2035–2039

    2018

  12. [20]

    M. A. Ebied, A. Munshi, S. A. Alhuzali, M. M. El-Sotouhy, A. I. Shehta, M. Elborlsy, Advanced deep learning modeling to enhance detection of defective photovoltaic cells in electroluminescence images, Scientific Reports 15 (1) (2025) 31640. 32

    2025

  13. [21]

    A. Jha, Y. Rawat, S. Vyas, Advancing automatic photovoltaic defect detection using semi-supervised semantic segmentation of electrolumi- nescence images, Engineering Applications of Artificial Intelligence 160 (2025) 111790

    2025

  14. [22]

    T. Yu, R. Feng, R. Feng, J. Liu, X. Jin, W. Zeng, Z. Chen, Inpaint anything: Segment anything meets image inpainting, arXiv preprint arXiv:2304.06790 (2023)

    2023 arXiv

  15. [23]

    J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, L. Fei-Fei, Imagenet: A large-scale hierarchical image database, in: 2009 IEEE conference on computer vision and pattern recognition, Ieee, 2009, pp. 248–255

    2009

  16. [24]

    Dosovitskiy, L

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, et al., An image is worth 16x16 words: Transformers for image recognition at scale, arXiv preprint arXiv:2010.11929 (2020)

    2020 arXiv

  17. [25]

    Devlin, M.-W

    J. Devlin, M.-W. Chang, K. Lee, K. Toutanova, Bert: Pre-training of deep bidirectional transformers for language understanding, in: Pro- ceedings of the 2019 conference of the North American chapter of the association for computational linguistics: human language technologies,...

    2019

  18. [26]

    Steiner, A

    A. Steiner, A. Kolesnikov, X. Zhai, R. Wightman, J. Uszkoreit, L. Beyer, How to train your vit? data, augmentation, and regularization in vision transformers, arXiv preprint arXiv:2106.10270 (2021)

    2021 arXiv

  19. [27]

    M. Tan, Q. Le, Efficientnetv2: Smaller models and faster training, in: International conference on machine learning, PMLR, 2021, pp. 10096– 10106

    2021

  20. [28]

    L.-C. Chen, G. Papandreou, F. Schroff, H. Adam, Rethinking atrous convolution for semantic image segmentation, arXiv preprint arXiv:1706.05587 (2017)

    2017 arXiv

  21. [29]

    K. He, X. Zhang, S. Ren, J. Sun, Deep residual learning for image recognition, in: Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 770–778. 33

    2016

  22. [30]

    Abdelsattar, A

    M. Abdelsattar, A. Abdelmoety, M. A. Ismeil, A. Emad-Eldeen, Au- tomated defect detection in solar cell images using deep learning algo- rithms, IEEE Access (2025)

    2025

  23. [31]

    Sandler, A

    M. Sandler, A. Howard, M. Zhu, A. Zhmoginov, L.-C. Chen, Mo- bilenetv2: Inverted residuals and linear bottlenecks, in: Proceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 4510–4520

    2018

  24. [32]

    Chollet, Xception: Deep learning with depthwise separable convolu- tions, in: Proceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp

    F. Chollet, Xception: Deep learning with depthwise separable convolu- tions, in: Proceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 1251–1258

    2017

  25. [33]

    Ghiasi, Y

    G. Ghiasi, Y. Cui, A. Srinivas, R. Qian, T.-Y. Lin, E. D. Cubuk, Q. V. Le, B. Zoph, Simple copy-paste is a strong data augmentation method for instance segmentation, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2021, pp. 2918–2928

    2021

  26. [34]

    Kajari-Schröder, I

    S. Kajari-Schröder, I. Kunze, U. Eitner, M. Köntges, Spatial and orien- tational distribution of cracks in crystalline photovoltaic modules gen- erated by mechanical load tests, Solar energy materials and solar cells 95 (11) (2011) 3054–3059

    2011

  27. [35]

    Sachert, Investigations of different soldering failure modes and their impact on module reliability, Energy Procedia 55 (2014) 456–463

    M.Heimann, R.Bakowskie, M.Köhler, J.Hirsch, M.Junghänel, A.Hus- sack, S. Sachert, Investigations of different soldering failure modes and their impact on module reliability, Energy Procedia 55 (2014) 456–463

    2014

  28. [36]

    Hooper, M

    S. Hooper, M. Chen, K. Saab, K. Bhatia, C. Langlotz, C. Ré, A case for reframing automated medical image classification as segmentation, Advances in Neural Information Processing Systems 36 (2023) 55415– 55441. 34

    2023

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.