pith. sign in

arxiv: 2606.02532 · v1 · pith:JBIRUPKCnew · submitted 2026-06-01 · 💻 cs.CV

Improving Combined Detection and Classification of TEM Defects via Mask-Conditioned Latent Diffusion Augmentation

Ni Li , Nuohao Liu , Ryan Jacobs , Ajay Annamareddy , Maciej P. Polak , Kevin Field , Izabela Szlufarska , Dane Morgan This is my paper

Pith reviewed 2026-06-28 14:42 UTC · model grok-4.3

classification 💻 cs.CV
keywords TEM defectslatent diffusiondata augmentationMask R-CNNdefect detectiondefect classificationmicrostructural analysisgenerative augmentation
0
0 comments X

The pith

Augmenting small TEM datasets with mask-conditioned latent diffusion generated images improves combined defect detection and classification F1 by up to 0.02.

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

The paper tests whether synthetic TEM images can help train better models when real labeled data for microstructural defects is limited. A latent diffusion model is conditioned on defect masks to generate new image and label pairs automatically from patterns in experimental data. These synthetic pairs are mixed into training sets of 10, 50, or 100 real images to train a Mask R-CNN. The combined detection and classification performance shows small gains in the harmonic mean of F1 scores. Benefits vary depending on how the data is split between training and testing.

Core claim

Mask-conditioned latent diffusion models can produce realistic synthetic TEM images with automatic multi-class defect masks that, when used to augment small experimental datasets, yield up to a 0.02 improvement in the harmonic mean of detection and classification F1 scores for a Mask R-CNN model.

What carries the argument

Mask-conditioned latent diffusion model that learns distributions from experimental masks to generate controllable synthetic image-mask pairs for data augmentation.

If this is right

  • Generative augmentation provides small performance boosts on datasets as small as 10 labeled images.
  • The improvement in detection versus classification tasks varies with the specific train and test data split.
  • Generation requires no additional manual annotations beyond the initial experimental masks.
  • Targeted generative models can support deep learning in data-scarce microscopy quantification.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • This method could be tested on other types of microscopy images if mask patterns are similar.
  • Further work might check whether the gains hold when base datasets are larger than 100 images.
  • The approach reduces reliance on expert time for labeling new defect instances.

Load-bearing premise

The synthetic images match real TEM images closely enough that adding them improves performance on real test data rather than causing the model to learn synthetic-specific features.

What would settle it

If a model trained on the augmented data shows lower or equal harmonic mean F1 on real held-out test images compared to training on real data alone, the claimed improvement would not hold.

Figures

Figures reproduced from arXiv: 2606.02532 by Ajay Annamareddy, Dane Morgan, Izabela Szlufarska, Kevin Field, Maciej P. Polak, Ni Li, Nuohao Liu, Ryan Jacobs.

Figure 1
Figure 1. Figure 1: Mask-conditioned latent diffusion pipeline for STEM image generation. Simulated (SIM) masks are sampled [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Visual comparison between real (left) and generated data (right). Left: Real segmentation masks and [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Detection F1, classification F1, and harmonic F1 scores as functions of confidence threshold for Mask R-CNN [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Comparison of Mask R-CNN performance across two train/test splits (Split 1 and Split 2). Panels (a) and [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Detection and classification performance of Mask R-CNN models trained on experimental (EXP) data alone [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
read the original abstract

Analyzing microstructural defects in transmission electron microscopy (TEM) images, particularly in irradiated metal alloys, is often limited by the availability of high-quality, labeled data. To address this, we introduce a generative data augmentation approach using a mask-conditioned latent diffusion model (LDM) for synthesizing realistic TEM images with controllable, automatically labeled multi-class defect masks. Without requiring manual annotations for generation, our method enables the creation of synthetic image-mask pairs by sampling distributions learned from experimental masks. These generated data were used to augment small experimental datasets of varying sizes (10, 50, and 100 labeled experimental images) to train a Mask Regional Convolutional Neural Network (R-CNN) model for defect detection and classification. Our results show that generative augmentation yields small overall model performance improvements, with up to a 0.02 gain in the harmonic mean of detection and classification F1 scores. However, we also find that the relative contributions to detection and classification improvement depend on the specific train/test data split. These findings highlight the potential of targeted generative models to enhance deep learning performance in data-scarce microscopy-based image quantification tasks.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper introduces a mask-conditioned latent diffusion model (LDM) to synthesize realistic TEM images paired with multi-class defect masks from learned distributions of experimental masks. These synthetic pairs augment small real training sets (sizes 10/50/100) to improve Mask R-CNN performance on combined defect detection and classification. Results report modest gains, with a maximum 0.02 increase in the harmonic mean of detection and classification F1 scores, though relative contributions vary by train/test split.

Significance. If the reported gains prove robust and attributable to the augmentation rather than artifacts, the method offers a practical way to address data scarcity in TEM-based materials analysis without manual annotation for synthetics. The mask-conditioning approach enables controllable generation, which is well-suited to the domain. Concrete numeric results on experimental data are a strength, but the small magnitude and split dependence limit broader significance unless further controls are added.

major comments (2)
  1. [Abstract] Abstract (results paragraph): the claim that generative augmentation produces a genuine 0.02 harmonic-mean F1 lift is load-bearing for the central contribution, yet no quantitative distribution-similarity metric (FID, LPIPS, or defect-statistic match) between real and LDM-generated images is provided; without it the modest gain could arise from split-specific overlap rather than improved training data.
  2. [Results] Results (split-dependence discussion): the observation that relative detection vs. classification gains depend on the specific train/test split directly undermines robustness of the augmentation benefit; an ablation using deliberately mismatched generated masks (as suggested by the stress-test note) is required to rule out coincidental correlation with the chosen test distribution.
minor comments (1)
  1. [Abstract] Abstract: the phrase 'up to a 0.02 gain' should be accompanied by the exact baseline and augmented F1 values per split for immediate clarity.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments, which help clarify the robustness of our claims. We address each major comment below and have updated the manuscript accordingly.

read point-by-point responses

  1. Referee: [Abstract] Abstract (results paragraph): the claim that generative augmentation produces a genuine 0.02 harmonic-mean F1 lift is load-bearing for the central contribution, yet no quantitative distribution-similarity metric (FID, LPIPS, or defect-statistic match) between real and LDM-generated images is provided; without it the modest gain could arise from split-specific overlap rather than improved training data.

    Authors: We agree that the absence of explicit distribution-similarity metrics leaves open the possibility that gains arise from split-specific effects rather than data quality. In the revised manuscript we will add FID scores computed on real versus generated images, along with comparisons of defect size, density, and class distributions, to provide quantitative support for the claim. revision: yes

  2. Referee: [Results] Results (split-dependence discussion): the observation that relative detection vs. classification gains depend on the specific train/test split directly undermines robustness of the augmentation benefit; an ablation using deliberately mismatched generated masks (as suggested by the stress-test note) is required to rule out coincidental correlation with the chosen test distribution.

    Authors: The split dependence is already noted in the manuscript, but we concur that it weakens the robustness argument without further controls. We will therefore perform the requested ablation with deliberately mismatched generated masks and report the outcomes to demonstrate that performance improvements are not explained by coincidental alignment with the test distribution. revision: yes

Circularity Check

0 steps flagged

No circularity: purely empirical ML augmentation experiment

full rationale

The paper reports measured F1 improvements on held-out real test images after augmenting small real training sets with LDM-generated images. No derivation, equation, or first-principles claim is presented that reduces the reported performance numbers to a fitted parameter, self-citation, or input by construction. The central result is an empirical observation whose validity rests on the experimental protocol (train/test splits, model training) rather than any internal definitional loop. This matches the default case of a self-contained empirical study.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review; no explicit free parameters, axioms, or invented entities are stated. Standard assumptions of diffusion models and supervised detection are implicit but not detailed.

pith-pipeline@v0.9.1-grok · 5754 in / 1019 out tokens · 23510 ms · 2026-06-28T14:42:42.300372+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

49 extracted references · 37 canonical work pages

  1. [1]

    Was.Fundamentals of Radiation Materials Science: Metals and Alloys

    Gary S. Was.Fundamentals of Radiation Materials Science: Metals and Alloys. Springer, New York, NY , 2nd edition, 2016. ISBN 978-1-4939-3437-4. doi: 10.1007/978-1-4939-3438-1. URL https://doi.org/10.1007/ 978-1-4939-3438-1

    work page doi:10.1007/978-1-4939-3438-1 2016

  2. [2]

    Faster r-cnn: Towards real-time object detection with region proposal networks, 2016

    Shaoqing Ren, Kaiming He, Ross Girshick, and Jian Sun. Faster r-cnn: Towards real-time object detection with region proposal networks, 2016. arXiv:1506.01497

    Pith/arXiv arXiv 2016

  3. [3]

    Mask r-cnn, 2018

    Kaiming He, Georgia Gkioxari, Piotr Dollár, and Ross Girshick. Mask r-cnn, 2018. arXiv:1703.06870

    Pith/arXiv arXiv 2018

  4. [4]

    You only look once: Unified, real-time object detection, 2016

    Joseph Redmon, Santosh Divvala, Ross Girshick, and Ali Farhadi. You only look once: Unified, real-time object detection, 2016. arXiv:1506.02640

    Pith/arXiv arXiv 2016

  5. [5]

    Sundararaghavan, N

    Mingren Shen, Guanzhao Li, Dongxia Wu, Yuhan Liu, Jacob R.C. Greaves, Wei Hao, Nathaniel J. Krakauer, Leah Krudy, Jacob Perez, Varun Sreenivasan, Bryan Sanchez, Oigimer Torres-Velázquez, Wei Li, Kevin G. Field, and Dane Morgan. Multi defect detection and analysis of electron microscopy images with deep learning. Computational Materials Science, 199:110576...

    work page doi:10.1016/j.commatsci 2021

  6. [6]

    Greaves, Donglin Wang, Zeming Xie, Zitong Huang, Chao Wang, Kevin G

    Ryan Jacobs, Mingren Shen, Yuhan Liu, Wei Hao, Xiaoshan Li, Ruoyu He, Jacob R.C. Greaves, Donglin Wang, Zeming Xie, Zitong Huang, Chao Wang, Kevin G. Field, and Dane Morgan. Performance and limitations of deep learning semantic segmentation of multiple defects in transmission electron micrographs.Cell Reports Physical Science, 3(5):100876, 2022. ISSN 2666...

    work page doi:10.1016/j.xcrp.2022.100876 2022

  7. [7]

    Qinyun Chen, Chaohui Zheng, Yue Cui, Yan-Ru Lin, and Steven J. Zinkle. A deep learning model for automatic analysis of cavities in irradiated materials.Computational Materials Science, 221:112073, 2023. ISSN 0927-0256. doi: https://doi.org/10.1016/j.commatsci.2023.112073. URL https://www.sciencedirect.com/science/ article/pii/S0927025623000678

    work page doi:10.1016/j.commatsci.2023.112073 2023

  8. [8]

    Haley, Kevin G

    Mingren Shen, Guanzhao Li, Dongxia Wu, Yudai Yaguchi, Jack C. Haley, Kevin G. Field, and Dane Morgan. A deep learning based automatic defect analysis framework for in-situ tem ion irradiations.Computational Materials Science, 197:110560, 2021. ISSN 0927-0256. doi: https://doi.org/10.1016/j.commatsci.2021.110560. URL https://www.sciencedirect.com/science/a...

    work page doi:10.1016/j.commatsci.2021.110560 2021

  9. [9]

    Lynch, Steven Chen, Dane Morgan, and Kevin G

    Ryan Jacobs, Priyam Patki, Matthew J. Lynch, Steven Chen, Dane Morgan, and Kevin G. Field. Materials swelling revealed through automated semantic segmentation of cavities in electron microscopy images.Scientific Reports, 13(1):5178, Mar 2023. ISSN 2045-2322. doi: 10.1038/s41598-023-32454-2. URL https://doi.org/10. 1038/s41598-023-32454-2

    work page doi:10.1038/s41598-023-32454-2 2023

  10. [10]

    Deep learning object detection in materials science: Current state and future directions.Computa- tional Materials Science, 211:111527, 2022

    Ryan Jacobs. Deep learning object detection in materials science: Current state and future directions.Computa- tional Materials Science, 211:111527, 2022. ISSN 0927-0256. doi: https://doi.org/10.1016/j.commatsci.2022. 111527. URLhttps://www.sciencedirect.com/science/article/pii/S0927025622002804

    work page doi:10.1016/j.commatsci.2022 2022

  11. [11]

    Kevin G Field, Priyam Patki, Nasir Sharaf, Kai Sun, Laura Hawkins, Matthew Lynch, Ryan Jacobs, Dane D Morgan, Lingfeng He, and Christopher R Field. Real-time, on-microscope automated quantification of features in microcopy experiments using machine learning and edge computing.Microscopy and Microanalysis, 28(S1): 2046–2048, 08 2022. ISSN 1431-9276. doi: 1...

    work page doi:10.1017/s1431927622007929 2046

  12. [12]

    K. G. Field, R. Jacobs, Shen Mingen, M. Lynch, P. Patki, C. Field, and D. Morgan. Development and deployment of automated machine learning detection in electron microcopy experiments.Microscopy and Microanalysis, 27 (suppl.1):2136–7, 2021. ISSN 1431-9276. doi: 10.1017/S1431927621007704

    work page doi:10.1017/s1431927621007704 2021

  13. [13]

    and Morgan, Dane , title=

    Wei Li, Kevin G. Field, and Dane Morgan. Automated defect analysis in electron microscopic images.npj Computational Materials, 4(1):36, Jul 2018. ISSN 2057-3960. doi: 10.1038/s41524-018-0093-8. URL https: //doi.org/10.1038/s41524-018-0093-8. 10 Running Title for Header

    work page doi:10.1038/s41524-018-0093-8 2018

  14. [14]

    Holm, Ryan Cohn, Nan Gao, Andrew R

    Elizabeth A. Holm, Ryan Cohn, Nan Gao, Andrew R. Kitahara, Thomas P. Matson, Bo Lei, and Srujana Rao Yarasi. Overview: Computer vision and machine learning for microstructural characterization and analysis.Metallurgical and Materials Transactions A, 51(12):5985–5999, Dec 2020. ISSN 1543-1940. doi: 10.1007/s11661-020-06008-4. URLhttps://doi.org/10.1007/s11...

    work page doi:10.1007/s11661-020-06008-4 2020

  15. [15]

    Learning-based defect recognition for quasi-periodic hrstem images.Micron, 146:103069, 2021

    Nik Dennler, Antonio Foncubierta-Rodriguez, Titus Neupert, and Marilyne Sousa. Learning-based defect recognition for quasi-periodic hrstem images.Micron, 146:103069, 2021. ISSN 0968-4328. doi: https: //doi.org/10.1016/j.micron.2021.103069. URL https://www.sciencedirect.com/science/article/pii/ S0968432821000603

    work page doi:10.1016/j.micron.2021.103069 2021

  16. [16]

    Automatic detection, localization and segmentation of nano-particles with deep learning in microscopy images.Micron, 120:113–119, 2019

    Ayse Betul Oktay and Anıl Gurses. Automatic detection, localization and segmentation of nano-particles with deep learning in microscopy images.Micron, 120:113–119, 2019. ISSN 0968-4328. doi: https: //doi.org/10.1016/j.micron.2019.02.009. URL https://www.sciencedirect.com/science/article/pii/ S0968432818304013

    work page doi:10.1016/j.micron.2019.02.009 2019

  17. [17]

    John Olamofe, Lijun Qian, and Kevin G. Field. Performance evaluation of image super-resolution for cavity detection in irradiated materials.IEEE Access, 13:68052–68065, 2025. doi: 10.1109/ACCESS.2025.3559443

    work page doi:10.1109/access.2025.3559443 2025

  18. [18]

    Baderot, M

    J. Baderot, M. Grould, D. Misra, N. Clément, A. Hallal, S. Martinez, and J. Foucher. Application of deep-learning based techniques for automatic metrology on scanning and transmission electron microscopy images.Journal of Vacuum Science and Technology B, 40(5):054003, 09 2022. ISSN 2166-2746. doi: 10.1116/6.0001988. URL https://doi.org/10.1116/6.0001988

    work page doi:10.1116/6.0001988 2022

  19. [19]

    Quantifying dislocation-type defects in post irradiation examination via transfer learning.Scientific Reports, 15(1):15889, May 2025

    Michael Wu, Jeremy Sharapov, Matthew Anderson, Yu Lu, and Yaqiao Wu. Quantifying dislocation-type defects in post irradiation examination via transfer learning.Scientific Reports, 15(1):15889, May 2025. ISSN 2045-2322. doi: 10.1038/s41598-025-00238-5. URLhttps://doi.org/10.1038/s41598-025-00238-5

    work page doi:10.1038/s41598-025-00238-5 2025

  20. [20]

    A new paradigm in electron microscopy: Automated microstructure analysis utilizing a dynamic segmentation convolutional neutral network.Materials Today Advances, 21:100468,

    Stephen Taller, Luke Scime, and Ty Austin. A new paradigm in electron microscopy: Automated microstructure analysis utilizing a dynamic segmentation convolutional neutral network.Materials Today Advances, 21:100468,

  21. [21]

    doi: https://doi.org/10.1016/j.mtadv.2024.100468

    ISSN 2590-0498. doi: https://doi.org/10.1016/j.mtadv.2024.100468. URL https://www.sciencedirect. com/science/article/pii/S2590049824000055

    work page doi:10.1016/j.mtadv.2024.100468 2024

  22. [22]

    Patrick, Christopher R

    Matthew J. Patrick, Christopher R. Field, Lauren H. L. Grae, Jeffrey M. Rickman, Kevin G. Field, and Katayun Barmak. A comparative analysis of yolov8 and u-net image segmentation approaches for transmission electron micrographs of polycrystalline thin films.APL Machine Learning, 3(3):036105, 07 2025. ISSN 2770-9019. doi: 10.1063/5.0274266. URLhttps://doi....

    work page doi:10.1063/5.0274266 2025

  23. [23]

    Evaluation of Human- Bias in Labeling of Ambiguous Features in Electron Microscopy Machine Learning Models.Microscopy and Microanalysis, 29(Supplement_1):1493–1494, 07 2023

    Gabriella Bruno, Matthew J Lynch, Ryan Jacobs, Dane D Morgan, and Kevin G Field. Evaluation of Human- Bias in Labeling of Ambiguous Features in Electron Microscopy Machine Learning Models.Microscopy and Microanalysis, 29(Supplement_1):1493–1494, 07 2023. ISSN 1431-9276. doi: 10.1093/micmic/ozad067.767. URLhttps://doi.org/10.1093/micmic/ozad067.767

    work page doi:10.1093/micmic/ozad067.767 2023

  24. [24]

    Uberuaga, Cheng Sun, and Ju Li

    Dane Morgan, Ghanshyam Pilania, Adrien Couet, Blas P. Uberuaga, Cheng Sun, and Ju Li. Machine learning in nuclear materials research.Current Opinion in Solid State and Materials Science, 26(2):100975, 2022. ISSN 1359-0286. doi: https://doi.org/10.1016/j.cossms.2021.100975. URL https://www.sciencedirect.com/ science/article/pii/S1359028621000784

    work page doi:10.1016/j.cossms.2021.100975 2022

  25. [25]

    Santos, Yuting Luo, Sarbajit Banerjee, and Bai-Xiang Xu

    Binbin Lin, Nima Emami, David A. Santos, Yuting Luo, Sarbajit Banerjee, and Bai-Xiang Xu. A deep learned nanowire segmentation model using synthetic data augmentation.npj Computational Materials, 8 (1):88, Apr 2022. ISSN 2057-3960. doi: 10.1038/s41524-022-00767-x. URL https://doi.org/10.1038/ s41524-022-00767-x

    work page doi:10.1038/s41524-022-00767-x 2022

  26. [26]

    M. J. Lynch, R. Jacobs, G. A. Bruno, P. Patki, D. Morgan, and K. G. Field. Accelerating domain-aware electron microscopy analysis using deep learning models with synthetic data and image-wide confidence scoring.npj Computational Materials, 11(1):261, Aug 2025. ISSN 2057-3960. doi: 10.1038/s41524-025-01756-6. URL https://doi.org/10.1038/s41524-025-01756-6

    work page doi:10.1038/s41524-025-01756-6 2025

  27. [27]

    Antón Cid-Mejías, Raúl Alonso-Calvo, Helena Gavilán, José Crespo, and Víctor Maojo. A deep learning approach using synthetic images for segmenting and estimating 3d orientation of nanoparticles in em images.Computer Methods and Programs in Biomedicine, 202:105958, 2021. ISSN 0169-2607. doi: https://doi.org/10.1016/j.cmpb. 2021.105958. URLhttps://www.scien...

    work page doi:10.1016/j.cmpb 2021

  28. [28]

    Dauphin, and David Lopez-Paz

    Hongyi Zhang, Moustapha Cisse, Yann N. Dauphin, and David Lopez-Paz. mixup: Beyond empirical risk minimization, 2018. URLhttps://arxiv.org/abs/1710.09412. arXiv:1710.09412

    Pith/arXiv arXiv 2018

  29. [29]

    Generative adversarial networks.Commun

    Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial networks.Commun. ACM, 63(11):139–144, October 2020. ISSN 0001-0782. doi: 10.1145/3422622. URLhttps://doi.org/10.1145/3422622. 11 Running Title for Header

    work page doi:10.1145/3422622 2020

  30. [30]

    Diffusion models beat gans on image synthesis

    Prafulla Dhariwal and Alexander Nichol. Diffusion models beat gans on image synthesis. In M. Ranzato, A. Beygelzimer, Y . Dauphin, P.S. Liang, and J. Wortman Vaughan, editors,Advances in Neural Information Processing Systems, volume 34, pages 8780–8794. Curran Associates, Inc., 2021. URL https://proceedings. neurips.cc/paper_files/paper/2021/file/49ad23d1...

    2021

  31. [31]

    A review on generative adversarial networks: Algorithms, theory, and applications.IEEE Transactions on Knowledge and Data Engineering, 35(4):3313–3332,

    Jie Gui, Zhenan Sun, Yonggang Wen, Dacheng Tao, and Jieping Ye. A review on generative adversarial networks: Algorithms, theory, and applications.IEEE Transactions on Knowledge and Data Engineering, 35(4):3313–3332,

  32. [32]

    doi: 10.1109/TKDE.2021.3130191

    work page doi:10.1109/tkde.2021.3130191 2021

  33. [33]

    Data augmentation in microscopic images for material data mining.npj Computational Materials, 6(1): 125, Aug 2020

    Boyuan Ma, Xiaoyan Wei, Chuni Liu, Xiaojuan Ban, Haiyou Huang, Hao Wang, Weihua Xue, Stephen Wu, Mingfei Gao, Qing Shen, Michele Mukeshimana, Adnan Omer Abuassba, Haokai Shen, and Yanjing Su. Data augmentation in microscopic images for material data mining.npj Computational Materials, 6(1): 125, Aug 2020. ISSN 2057-3960. doi: 10.1038/s41524-020-00392-6. U...

    work page doi:10.1038/s41524-020-00392-6 2020

  34. [34]

    Systematic review of generative adversarial networks (gans) in cell microscopy: Trends, practices, and impact on image augmentation.PLOS ONE, 20(6):1–35, 06 2025

    Duway Nicolas Lesmes-Leon, Andreas Dengel, and Sheraz Ahmed. Systematic review of generative adversarial networks (gans) in cell microscopy: Trends, practices, and impact on image augmentation.PLOS ONE, 20(6):1–35, 06 2025. doi: 10.1371/journal.pone.0291217. URLhttps://doi.org/10.1371/journal.pone.0291217

    work page doi:10.1371/journal.pone.0291217 2025

  35. [35]

    Chunguang Shen, Jingxuan Zhao, Minghao Huang, Chenchong Wang, Yuqi Zhang, Wei Xu, and Shijian Zheng. Generation of micrograph-annotation pairs for steel microstructure recognition using the hybrid deep generative model in the case of an extremely small and imbalanced dataset.Materials Characterization, 217:114407, 2024. ISSN 1044-5803. doi: https://doi.or...

    work page doi:10.1016/j.matchar.2024.114407 2024

  36. [36]

    Deep generative models-assisted automated labeling for electron microscopy images segmentation, 2024

    Wenhao Yuan, Bingqing Yao, Shengdong Tan, Fengqi You, and Qian He. Deep generative models-assisted automated labeling for electron microscopy images segmentation, 2024. URL https://arxiv.org/abs/2407. 19544. arXiv:2407.19544

    arXiv 2024

  37. [37]

    High-resolution image synthesis with latent diffusion models, 2022

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. High-resolution image synthesis with latent diffusion models, 2022. URL https://arxiv.org/abs/2112.10752. arXiv:2112.10752

    Pith/arXiv arXiv 2022

  38. [38]

    Diffusion models in medical imaging: A comprehensive survey.Medical Im- age Analysis, 88:102846, 2023

    Amirhossein Kazerouni, Ehsan Khodapanah Aghdam, Moein Heidari, Reza Azad, Mohsen Fayyaz, Ilker Haci- haliloglu, and Dorit Merhof. Diffusion models in medical imaging: A comprehensive survey.Medical Im- age Analysis, 88:102846, 2023. ISSN 1361-8415. doi: https://doi.org/10.1016/j.media.2023.102846. URL https://www.sciencedirect.com/science/article/pii/S136...

    work page doi:10.1016/j.media.2023.102846 2023

  39. [39]

    Diffusion-based deep learning method for augmenting ultrastructural imaging and volume electron microscopy.Nature Communications, 15(1):4677, Jun 2024

    Chixiang Lu, Kai Chen, Heng Qiu, Xiaojun Chen, Gu Chen, Xiaojuan Qi, and Haibo Jiang. Diffusion-based deep learning method for augmenting ultrastructural imaging and volume electron microscopy.Nature Communications, 15(1):4677, Jun 2024. ISSN 2041-1723. doi: 10.1038/s41467-024-49125-z. URL https://doi.org/10.1038/ s41467-024-49125-z

    work page doi:10.1038/s41467-024-49125-z 2024

  40. [40]

    Riegler, and Vajira Thambawita

    Roman Macháˇcek, Leila Mozaffari, Zahra Sepasdar, Sravanthi Parasa, Pål Halvorsen, Michael A. Riegler, and Vajira Thambawita. Mask-conditioned latent diffusion for generating gastrointestinal polyp images. InProceedings of the 4th ACM Workshop on Intelligent Cross-Data Analysis and Retrieval, ICDAR ’23, page 1–9, New York, NY , USA, 2023. Association for ...

    work page doi:10.1145/3592571.3592978 2023

  41. [41]

    Field, Xunxiang Hu, Kenneth C

    Kevin G. Field, Xunxiang Hu, Kenneth C. Littrell, Yukinori Yamamoto, and Lance L. Snead. Radiation tolerance of neutron-irradiated model fe–cr–al alloys.Journal of Nuclear Materials, 465:746–755, 2015. ISSN 0022-3115. doi: https://doi.org/10.1016/j.jnucmat.2015.06.023. URL https://www.sciencedirect.com/science/article/ pii/S0022311515300489

    work page doi:10.1016/j.jnucmat.2015.06.023 2015

  42. [42]

    Field, Samuel A

    Kevin G. Field, Samuel A. Briggs, Xunxiang Hu, Yukinori Yamamoto, Richard H. Howard, and Kumar Sridharan. Heterogeneous dislocation loop formation near grain boundaries in a neutron-irradiated commercial fecral alloy. Journal of Nuclear Materials, 483:54–61, 2017. ISSN 0022-3115. doi: https://doi.org/10.1016/j.jnucmat.2016.10

    work page doi:10.1016/j.jnucmat.2016.10 2017

  43. [43]

    URLhttps://www.sciencedirect.com/science/article/pii/S0022311516306596

  44. [44]

    Field, Samuel A

    Kevin G. Field, Samuel A. Briggs, Kumar Sridharan, Yukinori Yamamoto, and Richard H. Howard. Dislocation loop formation in model fecral alloys after neutron irradiation below 1 dpa.Journal of Nuclear Materials, 495:20–26, 2017. ISSN 0022-3115. doi: https://doi.org/10.1016/j.jnucmat.2017.07.061. URL https://www. sciencedirect.com/science/article/pii/S00223...

    work page doi:10.1016/j.jnucmat.2017.07.061 2017

  45. [45]

    Parish, Kevin G

    Chad M. Parish, Kevin G. Field, Alicia G. Certain, and Janelle P. Wharry. Application of stem characterization for investigating radiation effects in bcc fe-based alloys.Journal of Materials Research, 30(9):1275–1289, May 2015. ISSN 2044-5326. doi: 10.1557/jmr.2015.32. URLhttps://doi.org/10.1557/jmr.2015.32. 12 Running Title for Header

    work page doi:10.1557/jmr.2015.32 2015

  46. [46]

    Yao, D.J

    B. Yao, D.J. Edwards, and R.J. Kurtz. Tem characterization of dislocation loops in irradiated bcc fe-based steels. Journal of Nuclear Materials, 434(1):402–410, 2013. ISSN 0022-3115. doi: https://doi.org/10.1016/j.jnucmat. 2012.12.002. URL https://www.sciencedirect.com/science/article/pii/S0022311512006587. Spe- cial Section on Spent Nuclear Fuel

    work page doi:10.1016/j.jnucmat 2013

  47. [47]

    Topacio, Najmeh Mashhadi, and S

    Abolfazl Zargari, Benjamin R. Topacio, Najmeh Mashhadi, and S. Ali Shariati. Enhanced cell segmentation with limited training datasets using cycle generative adversarial networks.iScience, 27(5):109740, 2024. ISSN 2589-0042. doi: 10.1016/j.isci.2024.109740. URLhttps://doi.org/10.1016/j.isci.2024.109740

    work page doi:10.1016/j.isci.2024.109740 2024

  48. [48]

    Few-shot image generation with diffusion models,

    Jingyuan Zhu, Huimin Ma, Jiansheng Chen, and Jian Yuan. Few-shot image generation with diffusion models,

  49. [49]

    arXiv:2211.03264

    URLhttps://arxiv.org/abs/2211.03264. arXiv:2211.03264. 13

    arXiv