Pith. sign in

REVIEW 3 major objections 5 minor 69 references

Deep Learning-based Multi Project InP Wafer Simulation for Unsupervised Surface Defect Detection

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read Defect-free 'golden die' simulated from CAD plans by a U-Net

desk verdict A credible applied U-Net for simulating defect-free InP wafers from CAD, but the core claim of defect-free output is not actually measured with the pixel-perfect labels the authors made. read the letter →

arxiv 2506.10713 v1 pith:BXDP34NX submitted 2025-06-12 cs.CV cs.AIeess.IV

classification cs.CVcs.AIeess.IV
keywords visualinspectiondefectdetectionmachinelearningneuralnetworksgoldenstandardsimulationindiumphosphidephotonicsmulti-projectwafer
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

This paper tries to establish that a U-Net trained on pairs of CAD layer bitmaps and photographs of defective wafers can synthesize a defect-free 'golden' wafer image, which then serves as a template for automated surface defect detection. In multi-project InP wafer runs, golden samples are usually unavailable because each design is produced only once, so manual inspection is the norm. The authors show that their deep-learning simulator outperforms a decision-tree baseline and that a pixelwise comparison between the simulated golden die and the actual wafer photo localizes defects, with near-perfect average precision on synthetic data. If true, this would give foundries a label-free way to build golden standards from manufacturing plans.

What carries the argument

The central object is the U-Net (an encoder-decoder fully convolutional network with skip connections) that maps a stack of binary CAD layer bitmaps to either RGB pixel values (regression) or per-pixel class scores over a 64-color quantized palette (classification). The skip connections align input and output structures, letting the network transfer CAD geometry into the photograph domain; the training objective is a dissimilarity measure between the simulated and real patches, and the same measures (L2, LPIPS, etc.) are later used as similarity metrics to generate defect score maps. The paper also uses variance-thresholded patch sampling to focus training on informative regions, and a per-wafer model because each wafer run has distinct visual characteristics.

What would settle it

Train the same U-Net on a synthetic wafer with pixel-perfect defect labels and deliberately high defect density covering a large fraction of the CAD features; if the simulator then reconstructs the defects or fails to localize them in the score map, the sparsity assumption is broken. Alternatively, inspect simulations on real wafers for hallucinated structures not present in the CAD layers, which the paper itself already observes in some LPIPS-trained models.

Watch

Extended reading notes

Core claim

The central claim is that a fully convolutional U-Net, fed with binary CAD manufacturing layers, can predict a photorealistic, defect-free wafer photograph even when the training target photographs contain unlabeled defects. Because defects are sparse relative to CAD-correlated structures, the network learns the regular process appearance rather than copying the anomalies, and the resulting simulation can be used as a golden standard for template-matching defect detection. The paper evaluates four training objectives—regression with L2 and LPIPS, classification with cross-entropy and focal loss—and finds all outperform a decision-tree baseline, with semantic segmentation on quantized color palettes giving the most consistent results. The authors report that the defect-free property holds even for a wafer with unusually high defect density, and that the generated golden die enables defect localization via pixelwise similarity scores, though simulation hallucinations and CAD-to-photo misalignment remain failure modes.

Load-bearing premise

The load-bearing premise is that manufacturing defects are rare enough and visually separable enough from the CAD-defined structures that the U-Net will learn the clean wafer appearance rather than reproducing the defects.

Editorial extensions

If this is right

  • Golden standards can be generated from CAD plans for multi-project wafer runs where no physical defect-free die exists.
  • Defect detection becomes possible on any user-defined wafer region without any labeled defect data.
  • The same simulation approach could monitor intermediate manufacturing steps, since CAD layers correspond to individual process steps.
  • Training is fast relative to the months-long manufacturing cycle, so the pipeline could run on modest or embedded hardware in a foundry.
  • Dissimilarity maps between simulation and photo can guide human annotation, cutting the roughly 20 hours per wafer that manual inspection currently requires.

Reading between the lines

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

  • Editorial inference: the sparsity assumption could be tested directly on the released synthetic data by training on high-defect-density variants and checking whether the simulator systematically reproduces defects, an experiment the paper does not run.
  • Editorial inference: the method likely transfers to other structured-manufacturing domains where CAD or layout plans coexist with surface images, such as PCB or MEMS fabrication, because the same sparse-anomaly-plus-strong-structural-prior condition holds there.
  • Editorial inference: using a perceptual loss like LPIPS improves simulated noise realism but may hurt defect detection because the photo noise is reproduced as structure; a task-specific objective that penalizes structure hallucination could be more useful than purely perceptual fidelity.
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

3 major / 5 minor

Summary. The paper proposes a U-Net-based method that takes CAD layer bitmaps as input and predicts wafer photographs, with the goal of producing a defect-free "simulated golden die" for template-matching-based defect detection on InP multi-project wafers. The authors train classification and regression variants with four training objectives on five synthetic datasets and nine real wafer datasets, compare them against decision-tree baselines, and evaluate simulation quality with similarity metrics in RGB and quantized color space. They also demonstrate a defect-detection workflow using pixelwise dissimilarity score maps. The central claim is that models trained on defective wafer photographs yield defect-free simulations that can serve as golden standards for unsupervised inspection.

Significance. If the central claim were established, the method would be practically valuable for InP multi-project wafer manufacturing, where golden standards are typically unavailable and manual inspection is costly. The paper has notable strengths: it introduces synthetic wafer datasets with pixel-perfect defect labels, trains a large set of models (167 models) across four objectives, provides extensive similarity-metric tables and correlation analyses, and honestly reports failure modes such as hallucinations. However, the key claim that the simulator removes defects rather than memorizing them is not quantitatively supported. All reported similarity metrics compare simulations to the defective target photographs, and the available pixel-perfect labels on the synthetic datasets are never used for the load-bearing validation. The result is therefore promising but not yet demonstrated.

major comments (3)
  1. [Section 3.2, Section 3.3, Appendix A] The quantitative evidence for defect-free simulation is missing. All reported metrics (L2, LPIPS, PSNR, cross-entropy, k-off) compare the simulation to the actual wafer photograph, which contains defects, and the "best" checkpoints are selected by L2 and LPIPS against that same defective ground truth (Section 3.2). A model that faithfully reproduces defects would receive good scores under this protocol, and Section 3.3 explicitly notes that models tend to rank better when evaluated with their own training objective. The pixel-perfect defect labels on S1-S5 are never used to measure whether defects are removed. I request a defect-reproduction analysis on the synthetic datasets: for example, report the fraction of labeled defect pixels that produce a high dissimilarity in the simulation, or the precision/recall of the simulation-vs-photo difference at the labeled defect masks, separately from non-defective background. This is the load-bearing test for the Conclusion's claim that "even when utilizing defective wafers as simulation targets, the trained models yield defect-free simulations."
  2. [Section 4] The defect-detection results are anecdotal. Only two patches are quantified (AP 0.98 in Figure 20 and AP 0.44 in Figure 21), and no aggregate average precision is reported over the validation split or across the five synthetic datasets with pixel-perfect labels. The statement that "automated defect detection by template matching in synthetic data is virtually perfect when the simulation is good enough" is therefore not supported by the presented evidence. Please report the mean and standard deviation of AP over all validation patches and datasets, ideally per defect class, and state the exact thresholding protocol used to binarize the score map.
  3. [Section 4, Figure 21] Hallucinations are reported as a known failure mode, but their frequency and impact are not quantified. The paper states that defect detection "fails completely when there is an anomaly in the simulated patch" and shows hallucinations generating false positives; this is the opposite failure mode of a defect-free golden die and directly threatens the central premise. Please quantify hallucination incidence (for example, the fraction of simulated patches containing artifacts and the total hallucinated area) and report detection AP both including and excluding hallucinated patches, or propose and evaluate a post-processing mechanism that identifies and rejects such patches.
minor comments (5)
  1. [Section 3.1 vs. Appendix B] The training configuration is internally inconsistent: Section 3.1 specifies SGD with an initial learning rate of 5e-3 and a multiplicative decay schedule, while Appendix B states that the optimizer was Adam with learning rate 1e-4 (beta1=0.9, beta2=0.999) and that a batch size of 32 was chosen, whereas Section 3.1 says batches have at most 128 patches. Please specify the exact optimizer, learning-rate schedule, batch size, and variance threshold used for each reported model so that the 167 models can be reproduced.
  2. [Section 1] The sentence "In Section 4 outlines a pathway for template matching-based defect detection using our simulations" is ungrammatical; it should read "Section 4 outlines" or "In Section 4 we outline".
  3. [Section 3.1] "We disconsider samples" is not standard English; use "discard" or "exclude".
  4. [Section 2] The sentence describing the synthetic datasets says they are "square images of 10k×10k pixels," but Figure D.25 shows a sample described as 2000×2667 pixels; please clarify whether the 10k×10k size refers to the full wafer image from which patches are extracted.
  5. [Appendix B] The variance-threshold values are reported only for a subset of models ("R2 on LPIPS (v=20)", "R3, R5 and R8 on LPIPS (v=0, 30 epochs)", etc.), but not for all models; a complete table of hyperparameters per dataset and objective would improve reproducibility.

Circularity Check

1 steps flagged · score 3.0 of 10

Mild circularity from checkpoint selection by the same L2/LPIPS metrics later reported as evidence; the central defect-free claim rests on an unverified expectation rather than a circular derivation.

  1. fitted input called prediction [Section 3.2 (Evaluation methodology) and Section 3.3 (Results)]
    "We present results for the models generated after the 10-th training epoch and for the best performing intermediate checkpoints. We select a set of best models according to the MSE similarity to the ground truth and a set of checkpoints based on their LPIPS performance."

    The 'best' checkpoint is selected on the validation split using the same L2/LPIPS metrics that are subsequently reported as evidence of simulation quality and used to rank models against the decision-tree baseline. By construction, the reported 'best epoch' L2/LPIPS value is the minimum over the evaluated checkpoints for that metric, while the decision tree is given no equivalent epoch-selection step; hence the claimed U-Net outperformance is partly a restatement of the selection rule rather than an independent prediction. The paper itself concedes that 'models tend to perform better when evaluated on the same function used as a training objective,' confirming that the metric ranking is confounded with the selection criterion.

full rationale

The core pipeline is not circular: the U-Net is trained to map CAD layer bitmaps to photographed wafer appearance, and the decision-tree baseline is a legitimate comparator apart from the checkpoint-selection bias. No load-bearing self-citation or imported-uniqueness argument appears; prior theses by the authors are cited only for alignment, quantization, and annotation details. The central weakness is that the 'defect-free golden die' claim is supported by qualitative samples and by similarity scores computed against defective ground-truth images, which cannot by themselves establish that defects were removed rather than memorized; this is a validity gap, not a definitional or self-citational circularity. The one genuine circular element is the overlap between the L2/LPIPS checkpoint-selection criterion and the L2/LPIPS evaluation, which is partial, acknowledged by the authors, and does not force the central conclusion by itself.

Assumptions & free parameters 5 free parameters · 3 assumptions · 0 invented entities

The central claim rests on the tunable palette size, per-dataset variance thresholds, checkpoint selection, and the unverified assumption that defects are too sparse to be learned. No new physical entities are introduced; the 'simulated golden die' is a method output, not a postulated entity.

free parameters (5)
  • Color palette size k = 64
    Chosen empirically in Appendix C; the classification models output one of 64 quantized colors, and the paper notes k-off accuracy suffers from sub-optimal palette ordering.
  • Variance threshold v for patch filtering = 0 to 50 (per dataset)
    Patches below a variance threshold are discarded during training; the threshold is tuned separately for several models (Appendix B), constituting post hoc data selection.
  • Batch size = 32
    Chosen because 'perceptually better results were observed' (Appendix B).
  • Learning rate and optimizer = SGD 5e-3 with decay in Section 3.1; Adam 1e-4 in Appendix B
    The main text and appendix describe different optimizers and learning rates, making the exact training configuration ambiguous.
  • Best epoch checkpoint = 10th epoch or best validation L2/LPIPS
    Checkpoints are selected based on the same metrics used for evaluation, giving the models an advantage not granted to the decision-tree baseline.
assumptions (3)
  • domain assumption The visual appearance of the wafer is determined by the CAD layers, so a function g -> p exists and is learnable.
    Invoked in Section 3 as the mapping p-hat(R) = U(g(R)); the entire simulation approach presupposes that CAD layouts contain sufficient information to render the wafer image.
  • domain assumption Manufacturing defects are sparse enough that a capacity-limited network will not systematically reproduce them in its output.
    Stated in Section 3: 'Due to the sparsity of these defects in the wafer photographs ... it is expected that the simulations will not systematically replicate those defects and anomalies.' This is the load-bearing premise for defect-free simulation and is not verified systematically on the labeled synthetic data.
  • standard math The U-Net with three downsampling stages can represent the wafer appearance function at 64x64 patch resolution.
    Architecture choice from Ronneberger et al. [43], assumed sufficient without an expressiveness analysis.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Deep Learning-based Multi Project InP Wafer Simulation for Unsupervised Surface Defect Detection." pith.science (2026). https://pith.science/paper/BXDP34NX

@misc{pith2026250610713,
  author       = {Pith},
  title        = {Pith review of: Deep Learning-based Multi Project InP Wafer Simulation for Unsupervised Surface Defect Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BXDP34NX}},
  note         = {Machine review of arXiv:2506.10713}
}
read the original abstract

Quality management in semiconductor manufacturing often relies on template matching with known golden standards. For Indium-Phosphide (InP) multi-project wafer manufacturing, low production scale and high design variability lead to such golden standards being typically unavailable. Defect detection, in turn, is manual and labor-intensive. This work addresses this challenge by proposing a methodology to generate a synthetic golden standard using Deep Neural Networks, trained to simulate photo-realistic InP wafer images from CAD data. We evaluate various training objectives and assess the quality of the simulated images on both synthetic data and InP wafer photographs. Our deep-learning-based method outperforms a baseline decision-tree-based approach, enabling the use of a 'simulated golden die' from CAD plans in any user-defined region of a wafer for more efficient defect detection. We apply our method to a template matching procedure, to demonstrate its practical utility in surface defect detection.

Figures

Figures reproduced from arXiv: 2506.10713 by the authors.

Figure 1
Figure 1. An overview of our proposed “golden die” simulation methodology. (1) A dataset [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Visualizations of the “score maps” generated by different defect detection meth [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Quantized palette samples from a real wafer (3(a)) and the synthetic wafer [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (20 more)
Figure 4
Figure 4. Figure 4: Visual representation of various data types associated with a wafer. Defect labels [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: A sample pair of a single CAD layer and corresponding image in our synthetic [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: The 4 defect classes implemented in the toy data: (a) dust particles, at 10,000 [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: L2-norm validation loss for the models trained on synthetic data, grouped [PITH_FULL_IMAGE:figures/full_fig_p014_7.png]
Figure 8
Figure 8. Figure 8: LPIPS validation loss for the models trained on synthetic data, grouped by which [PITH_FULL_IMAGE:figures/full_fig_p015_8.png]
Figure 9
Figure 9. Figure 9: Samples of a simulation on synthetic data. The simulator is a regression model [PITH_FULL_IMAGE:figures/full_fig_p016_9.png]
Figure 10
Figure 10. Figure 10: Samples of the prediction on mock data, purely background. On this patch, [PITH_FULL_IMAGE:figures/full_fig_p016_10.png]
Figure 11
Figure 11. Figure 11: L2-norm validation loss for the models trained on real data. They are grouped [PITH_FULL_IMAGE:figures/full_fig_p017_11.png]
Figure 12
Figure 12. Figure 12: LPIPS validation loss for all of the models trained on real MPW wafers. They [PITH_FULL_IMAGE:figures/full_fig_p018_12.png]
Figure 13
Figure 13. Figure 13: Samples of the prediction on real data. The U-Net based simulation scored [PITH_FULL_IMAGE:figures/full_fig_p019_13.png]
Figure 14
Figure 14. Figure 14: Samples of the prediction on real data. Whole wafer: L2 0.009, LPIPS 0.52 [PITH_FULL_IMAGE:figures/full_fig_p019_14.png]
Figure 15
Figure 15. Figure 15: Simulations from models trained on the indicated target function. Some models [PITH_FULL_IMAGE:figures/full_fig_p020_15.png]
Figure 16
Figure 16. Figure 16: Appearance of the ripples observed. We show a background patch for ease [PITH_FULL_IMAGE:figures/full_fig_p020_16.png]
Figure 17
Figure 17. Figure 17: A target patch and simulation samples, all 128 [PITH_FULL_IMAGE:figures/full_fig_p021_17.png]
Figure 18
Figure 18. Figure 18: Color profile along a vertical path through the centerline of simulated patches [PITH_FULL_IMAGE:figures/full_fig_p022_18.png]
Figure 19
Figure 19. Figure 19: Detailed view of the same plot as in Figure 18, focusing on the area containing [PITH_FULL_IMAGE:figures/full_fig_p023_19.png]
Figure 20
Figure 20. Figure 20: The defect detection works considerably well on the synthetic datasets, as we [PITH_FULL_IMAGE:figures/full_fig_p023_20.png]
Figure 21
Figure 21. Figure 21: The defect detection fails completely when there is an anomaly in the simulated [PITH_FULL_IMAGE:figures/full_fig_p024_21.png]
Figure 22
Figure 22. Figure 22: Regression model trained with LPIPS loss on a real wafer. Due to the stitching [PITH_FULL_IMAGE:figures/full_fig_p024_22.png]
Figure 23
Figure 23. Figure 23: Classification model trained with focal loss, on a wafer with very high defect [PITH_FULL_IMAGE:figures/full_fig_p025_23.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

69 extracted references · 64 canonical work pages

  1. [1]

    Unsupervised pixel-wise hyperspectral anomaly detection via autoencoding adversarial networks

    Arisoy, S., Nasrabadi, N.M., Kayabol, K., 2021. Unsupervised pixel-wise hyperspectral anomaly detection via autoencoding adversarial networks. IEEE Geoscience and Remote Sensing Letters 19, 1–5

  2. [2]

    Robust Image Wafer Inspection, in: 2020 Tenth International Conference on Image Processing Theory, Tools and Appli- cations (IPTA), IEEE

    Barone, M., 2020. Robust Image Wafer Inspection, in: 2020 Tenth International Conference on Image Processing Theory, Tools and Appli- cations (IPTA), IEEE. pp. 1–6

  3. [3]

    Hyperspectral anomaly detection method based on auto-encoder, in: Image and Signal Processing for Remote Sensing XXI, Spie

    Bati, E., Çalışkan, A., Koz, A., Alatan, A.A., 2015. Hyperspectral anomaly detection method based on auto-encoder, in: Image and Signal Processing for Remote Sensing XXI, Spie. pp. 220–226

  4. [4]

    Automatic defect classification: status and industry trends, in: Integrated Circuit Metrol- ogy, Inspection, and Process Control IX, SPIE

    Bennett, M.H., Tobin Jr, K.W., Gleason, S.S., 1995. Automatic defect classification: status and industry trends, in: Integrated Circuit Metrol- ogy, Inspection, and Process Control IX, SPIE. pp. 210–220

  5. [5]

    Image-based surface defect detection using deep learning: A review

    Bhatt, P.M., Malhan, R.K., Rajendran, P., Shah, B.C., Thakar, S., Yoon, Y.J., Gupta, S.K., 2021. Image-based surface defect detection using deep learning: A review. Journal of Computing and Information Science in Engineering 21

  6. [6]

    Wafer Defect Inspection by Neural Analysis of Region Features

    Chang, C.Y., Li, C.H., Chang, Y.C., Jeng, M., 2011. Wafer Defect Inspection by Neural Analysis of Region Features. Journal of Intelligent Manufacturing 22, 953–964. 28

  7. [7]

    Application of two Hopfield Neural Networks for Automatic four-element LED Inspection

    Chang, C.Y., Li, C.H., Lin, S.Y., Jeng, M., 2009. Application of two Hopfield Neural Networks for Automatic four-element LED Inspection. IEEE Transactions on Systems, Man, and Cybernetics, Part C (Appli- cations and Reviews) 39, 352–365

  8. [8]

    Semi-supervised learning

    Chapelle, O., Scholkopf, B., Zien, A., 2006. Semi-supervised learning

Show all 69 references
  1. [9]

    A Neural-Network Approach to Recognize DefectSpatialPatterninSemiconductorFabrication

    Chen, F.L., Liu, S.F., 2000. A Neural-Network Approach to Recognize DefectSpatialPatterninSemiconductorFabrication. IEEEtransactions on semiconductor manufacturing 13, 366–373

  2. [10]

    A Neural-Network Approach for an Au- tomatic LED Inspection System

    Chen, W.C., Hsu, S.W., 2007. A Neural-Network Approach for an Au- tomatic LED Inspection System. Expert Systems with Applications 33, 531–537

  3. [11]

    A light-weighted cnn model for wafer structural defect detection

    Chen, X., Chen, J., Han, X., Zhao, C., Zhang, D., Zhu, K., Su, Y., 2020. A light-weighted cnn model for wafer structural defect detection. IEEE access 8, 24006–24018

  4. [12]

    Convolutional neu- ral network for wafer surface defect classification and the detection of unknown defect class

    Cheon, S., Lee, H., Kim, C.O., Lee, S.H., 2019. Convolutional neu- ral network for wafer surface defect classification and the detection of unknown defect class. IEEE Transactions on Semiconductor Manufac- turing 32, 163–170

  5. [13]

    Automated visual inspection: A survey

    Chin, R.T., Harlow, C.A., 1982. Automated visual inspection: A survey. IEEE transactions on pattern analysis and machine intelligence , 557– 573

  6. [14]

    Implementation of Reflected Light Die-to-Die Inspection and ReviewSmart to Improve 65nm DRAM Mask Fabrication, in: 25th Annual BACUS Symposium on Photomask Technology, SPIE

    Cho, W.I., Park, J.H., Chung, D.H., Cha, B.C., Choi, S.W., Han, W.S., Park, K.H., Kim, N.W., Hess, C., Ma, W., et al., 2005. Implementation of Reflected Light Die-to-Die Inspection and ReviewSmart to Improve 65nm DRAM Mask Fabrication, in: 25th Annual BACUS Symposium on Photom...

  7. [15]

    Chou, P.B., Rao, A.R., Sturzenbecker, M.C., Wu, F.Y., Brecher, V.H.,

  8. [16]

    Çiçek, Ö., Abdulkadir, A., Lienkamp, S.S., Brox, T., Ronneberger, O.,

  9. [17]

    An image-based pipeline for defect local- ization in photonic InP multi-project wafers

    Dolgener Cantú, E., 2023. An image-based pipeline for defect local- ization in photonic InP multi-project wafers. Master’s thesis. Tech- nische Universität Berlin. URL:https://doi.org/10.5281/zenodo. 15230127, doi:10.5281/zenodo.15230127

  10. [18]

    Recent advances in the automatic in- spection of integrated circuits for pattern defects

    Dom, B.E., Brecher, V., 1995. Recent advances in the automatic in- spection of integrated circuits for pattern defects. Machine Vision and Applications 8, 5–19

  11. [19]

    Age and gender estimation of unfiltered faces

    Eidinger, E., Enbar, R., Hassner, T., 2014. Age and gender estimation of unfiltered faces. IEEE Transactions on information forensics and security 9, 2170–2179

  12. [20]

    Unsuper- vised fabric defect detection based on a deep convolutional generative adversarial network

    Hu, G., Huang, J., Wang, Q., Li, J., Xu, Z., Huang, X., 2020. Unsuper- vised fabric defect detection based on a deep convolutional generative adversarial network. Textile Research Journal 90, 247–270

  13. [21]

    Runet: A robust unet architecture for image super-resolution, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recog- nition Workshops, pp

    Hu, X., Naiel, M.A., Wong, A., Lamm, M., Fieguth, P., 2019. Runet: A robust unet architecture for image super-resolution, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recog- nition Workshops, pp. 0–0

  14. [22]

    Huang, G., Laradji, I., Vazquez, D., Lacoste-Julien, S., Rodriguez, P.,

  15. [23]

    Automated visual inspection in the semi- conductor industry: A survey

    Huang, S.H., Pan, Y.C., 2015. Automated visual inspection in the semi- conductor industry: A survey. Computers in industry 66, 1–10

  16. [24]

    A cnn-based transfer learning method for defect classification in semiconductor man- ufacturing, in: 2018 international symposium on semiconductor manu- facturing (ISSM), IEEE

    Imoto, K., Nakai, T., Ike, T., Haruki, K., Sato, Y., 2018. A cnn-based transfer learning method for defect classification in semiconductor man- ufacturing, in: 2018 international symposium on semiconductor manu- facturing (ISSM), IEEE. pp. 1–3. 30

  17. [25]

    Batch normalization: Accelerating deep network training by reducing internal covariate shift, in: International conference on machine learning, pmlr

    Ioffe, S., Szegedy, C., 2015. Batch normalization: Accelerating deep network training by reducing internal covariate shift, in: International conference on machine learning, pmlr. pp. 448–456

  18. [26]

    Image-to-image transla- tion with conditional adversarial networks, in: Proceedings of the IEEE conference on computer vision and pattern recognition, pp

    Isola, P., Zhu, J.Y., Zhou, T., Efros, A.A., 2017. Image-to-image transla- tion with conditional adversarial networks, in: Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 1125–1134

  19. [27]

    Color image generation from lidar reflection data by using selected connection unet

    Kim, H.K., Yoo, K.Y., Jung, H.Y., 2020. Color image generation from lidar reflection data by using selected connection unet. Sensors 20, 3387

  20. [28]

    Novel method for detection of mixed-type defect patterns in wafer maps based on a single shot detector algorithm

    Kim, T.S., Lee, J.W., Lee, W.K., Sohn, S.Y., 2022. Novel method for detection of mixed-type defect patterns in wafer maps based on a single shot detector algorithm. Journal of Intelligent Manufacturing 33, 1715– 1724

  21. [29]

    System-on- chip photonic integrated circuits

    Kish, F., Lal, V., Evans, P., Corzine, S.W., Ziari, M., Butrie, T., Reffle, M., Tsai, H.S., Dentai, A., Pleumeekers, J., et al., 2017. System-on- chip photonic integrated circuits. IEEE Journal of Selected Topics in Quantum Electronics 24, 1–20

  22. [30]

    Kohlberger, T., Singh, V., Alvino, C., Bahlmann, C., Grady, L., 2012. Evaluating segmentation error without ground truth, in: Medical Image Computing and Computer-Assisted Intervention–MICCAI 2012: 15th International Conference, Nice, France, October 1-5, 2012, Proceedings, Pa...

  23. [31]

    Epitaxial growth and layer-transfer techniques for heterogeneous integration of materials for electronic and photonic devices

    Kum, H., Lee, D., Kong, W., Kim, H., Park, Y., Kim, Y., Baek, Y., Bae, S.H., Lee, K., Kim, J., 2019. Epitaxial growth and layer-transfer techniques for heterogeneous integration of materials for electronic and photonic devices. Nature Electronics 2, 439–450

  24. [32]

    Wavelet-baseddefectdetectioninsolarwafer images with inhomogeneous texture

    Li, W.C., Tsai, D.M., 2012. Wavelet-baseddefectdetectioninsolarwafer images with inhomogeneous texture. Pattern Recognition 45, 742–756

  25. [33]

    Concealed object segmenta- tion in terahertz imaging via adversarial learning

    Liang, D., Pan, J., Yu, Y., Zhou, H., 2019. Concealed object segmenta- tion in terahertz imaging via adversarial learning. Optik 185, 1104–1114

  26. [34]

    Foundry 31 capabilities for photonic integrated circuits, in: Optical Fiber Telecom- munications VII

    Liehr, M., Baier, M., Hoefler, G., Fahrenkopf, N.M., Bowers, J., Glad- hill, R., O’Brien, P., Timurdogan, E., Su, Z., Kish, F., 2020. Foundry 31 capabilities for photonic integrated circuits, in: Optical Fiber Telecom- munications VII. Elsevier, pp. 143–193

  27. [35]

    Focal loss for dense object detection, in: Proceedings of the IEEE international conference on computer vision, pp

    Lin, T.Y., Goyal, P., Girshick, R., He, K., Dollár, P., 2017. Focal loss for dense object detection, in: Proceedings of the IEEE international conference on computer vision, pp. 2980–2988

  28. [36]

    Defect detec- tion of ic wafer based on spectral subtraction

    Liu, H., Zhou, W., Kuang, Q., Cao, L., Gao, B., 2010. Defect detec- tion of ic wafer based on spectral subtraction. IEEE transactions on semiconductor manufacturing 23, 141–147

  29. [37]

    Deep learning for generic object detection: A survey

    Liu, L., Ouyang, W., Wang, X., Fieguth, P., Chen, J., Liu, X., Pietikäi- nen, M., 2020. Deep learning for generic object detection: A survey. International journal of computer vision 128, 261–318

  30. [38]

    Independent component analysis-based de- fect detection in patterned liquid crystal display surfaces

    Lu, C.J., Tsai, D.M., 2008. Independent component analysis-based de- fect detection in patterned liquid crystal display surfaces. Image and Vision Computing 26, 955–970

  31. [39]

    A review of automatic fabric defect detection techniques

    Mahajan, P., Kolhe, S., Patil, P., 2009. A review of automatic fabric defect detection techniques. Advances in Computational Research 1, 18–29

  32. [40]

    Unsupervised surface defect detection using deep autoencoders and data augmentation, in: 2018 International Conference on Cyberworlds (CW), IEEE

    Mujeeb, A., Dai, W., Erdt, M., Sourin, A., 2018. Unsupervised surface defect detection using deep autoencoders and data augmentation, in: 2018 International Conference on Cyberworlds (CW), IEEE. pp. 391– 398

  33. [41]

    Patterned fabric defect detection and classification (fddc) techniques: a review

    Oni, D., Ojo, J., Alabi, B., Adebayo, A., Amoran, A., 2018. Patterned fabric defect detection and classification (fddc) techniques: a review. International Journal of Scientific & Engineering Research 9, 1156–1165

  34. [42]

    Video generation from single semantic label map, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp

    Pan, J., Wang, C., Jia, X., Shao, J., Sheng, L., Yan, J., Wang, X., 2019. Video generation from single semantic label map, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 3733–3742

  35. [43]

    U-net: Convolutional networks for biomedical image segmentation.arXiv:1505.04597

    Ronneberger, O., Fischer, P., Brox, T., 2015. U-net: Convolutional networks for biomedical image segmentation.arXiv:1505.04597. 32

  36. [44]

    Unsupervised anomaly detection with generative adversar- ial networks to guide marker discovery, in: International conference on information processing in medical imaging, Springer

    Schlegl, T., Seeböck, P., Waldstein, S.M., Schmidt-Erfurth, U., Langs, G., 2017. Unsupervised anomaly detection with generative adversar- ial networks to guide marker discovery, in: International conference on information processing in medical imaging, Springer. pp. 146–157

  37. [45]

    Training region-based object detectors with online hard example mining, in: Proceedings of the IEEE conference on computer vision and pattern recognition, pp

    Shrivastava, A., Gupta, A., Girshick, R., 2016. Training region-based object detectors with online hard example mining, in: Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 761–769

  38. [46]

    Smit, M., Leijtens, X., Ambrosius, H., Bente, E., Van der Tol, J., Smal- brugge, B., De Vries, T., Geluk, E.J., Bolk, J., Van Veldhoven, R., et al.,

  39. [47]

    Past, present, and future of inp-based photonic integration

    Smit, M., Williams, K., Van Der Tol, J., 2019. Past, present, and future of inp-based photonic integration. APL Photonics 4, 050901

  40. [48]

    InP-based foundry PICs for optical interconnects

    Soares, F.M., Baier, M., Gaertner, T., Grote, N., Moehrle, M., Becker- werth, T., Runge, P., Schell, M., . InP-based foundry PICs for optical interconnects. Applied Sciences 9, 1588. URL:https://www.mdpi.com/ 2076-3417/9/8/1588, doi:10.3390/app9081588

  41. [49]

    A Neural-Network Approach for Semiconductor Wafer Post-sawing Inspection

    Su, C.T., Yang, T., Ke, C.M., 2002. A Neural-Network Approach for Semiconductor Wafer Post-sawing Inspection. IEEE Transactions on Semiconductor Manufacturing 15, 260–266

  42. [50]

    Segmentation- based deep-learning approach for surface-defect detection

    Tabernik, D., Šela, S., Skvarč, J., Skočaj, D., 2020. Segmentation- based deep-learning approach for surface-defect detection. Journal of Intelligent Manufacturing 31, 759–776

  43. [51]

    Background modelling based on generative unet, in: 2017 14th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS), IEEE

    Tao, Y., Palasek, P., Ling, Z., Patras, I., 2017. Background modelling based on generative unet, in: 2017 14th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS), IEEE. pp. 1–6

  44. [52]

    An eigenvalue-based similarity measure and its application in defect detection

    Tsai, D.M., Yang, R.H., 2005. An eigenvalue-based similarity measure and its application in defect detection. Image and Vision Computing 23, 1094–1101. 33

  45. [53]

    De- tection defect in printed circuit boards using unsupervised feature ex- traction upon transfer learning, in: 2019 International Conference on Cyberworlds (CW), IEEE

    Volkau, I., Mujeeb, A., Wenting, D., Marius, E., Alexei, S., 2019. De- tection defect in printed circuit boards using unsupervised feature ex- traction upon transfer learning, in: 2019 International Conference on Cyberworlds (CW), IEEE. pp. 101–108

  46. [54]

    Detection and classification of defect patterns on semiconductor wafers

    Wang, C.H., Kuo, W., Bensmail, H., 2006. Detection and classification of defect patterns on semiconductor wafers. IIE transactions 38, 1059– 1068

  47. [55]

    Wang, T.C., Liu, M.Y., Zhu, J.Y., Tao, A., Kautz, J., Catanzaro, B.,

  48. [56]

    A wafer surface defect detection method built on generic object detection network

    Wang, X., Jia, X., Jiang, C., Jiang, S., 2022. A wafer surface defect detection method built on generic object detection network. Digital Signal Processing , 103718

  49. [57]

    PhotorealisticSimulationofPhotonic-Integrated Circuit Designs using Digital Image Processing Techniques and Neural Networks

    Wittmann, R.K., 2022. PhotorealisticSimulationofPhotonic-Integrated Circuit Designs using Digital Image Processing Techniques and Neural Networks. Master’s thesis. Technische Universität Berlin

  50. [58]

    A review of recent advances in surface defect detection us- ing texture analysis techniques

    Xie, X., 2008. A review of recent advances in surface defect detection us- ing texture analysis techniques. ELCVIA: electronic letters on computer vision and image analysis , 1–22

  51. [59]

    Evaluating explanation without ground truthininterpretablemachinelearning

    Yang, F., Du, M., Hu, X., 2019. Evaluating explanation without ground truthininterpretablemachinelearning. arXivpreprintarXiv:1907.06831

  52. [60]

    Weakly supervised object localization and detection: A survey

    Zhang, D., Han, J., Cheng, G., Yang, M.H., 2021. Weakly supervised object localization and detection: A survey. IEEE transactions on pat- tern analysis and machine intelligence 44, 5866–5885

  53. [61]

    The development of an automatic post-sawing inspection system using computer vision tech- niques

    Zhang, J.M., Lin, R.M., Wang, M.J.J., 1999. The development of an automatic post-sawing inspection system using computer vision tech- niques. Computers in Industry 40, 51–60

  54. [62]

    The unreasonable effectiveness of deep features as a perceptual metric, in: 34 Proceedings of the IEEE conference on computer vision and pattern recognition, pp

    Zhang, R., Isola, P., Efros, A.A., Shechtman, E., Wang, O., 2018. The unreasonable effectiveness of deep features as a perceptual metric, in: 34 Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 586–595

  55. [63]

    Object detection in 20 years: A survey

    Zou, Z., Shi, Z., Guo, Y., Ye, J., 2019. Object detection in 20 years: A survey. arXiv preprint arXiv:1905.05055 . 35 Appendix A. Evaluation metrics and complete quantitative simi- larity results We perform quantitative evaluation of simulation quality by computation of a few ...

  56. [1997]

    Machine vision and applications 9, 201–214

    Automatic defect classification for semiconductor manufacturing. Machine vision and applications 9, 201–214

  57. [2006]

    Cambridge, Massachusettes: The MIT Press View Article 2

  58. [2014]

    Semi- conductor Science and Technology 29, 083001

    An introductionto inp-based genericintegration technology. Semi- conductor Science and Technology 29, 083001

  59. [2016]

    3d u-net: learning dense volumetric segmentation from sparse 29 annotation, in: Medical Image Computing and Computer-Assisted Intervention–MICCAI 2016: 19th International Conference, Athens, Greece, October 17-21, 2016, Proceedings, Part II 19, Springer. pp. 424– 432

  60. [2018]

    8798–8807

    High-resolution image synthesis and semantic manipulation with conditional gans, in: Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 8798–8807

  61. [2021]

    arXiv preprint arXiv:2110.14711

    A survey of self-supervised and few-shot object detection. arXiv preprint arXiv:2110.14711

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.