Pith. sign in

REVIEW 4 major objections 6 minor 12 references

A Framework for Supervised and Unsupervised Segmentation and Classification of Materials Microstructure Images

T0 review · 4 major / 6 minor · reviewed 2026-08-08 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a largely automated, iterative three-step pipeline can segment multiphase micrographs, classify regions into known or newly discovered microstructure classes, and build a database that improves its own performance.

desk verdict A real assembly job with useful pieces—but the paper validates components separately, never runs the advertised iterative loop on real materials, and skips the stationary single-phase control that would anchor Step 1. read the letter →

arxiv 2502.07107 v2 pith:CR7CD4SH submitted 2025-02-10 stat.AP cs.CVstat.ML

Kungang Zhang , Wei Chen , Wing K. Liu , L. Catherine Brinson , Daniel W. Apley This is my paper · ORCID
classification stat.APcs.CVstat.ML
keywords microstructurecharacterizationFisherscorevectorunsupervisedsegmentationevidentialdeeplearningsemantictransfermaterialsinformaticsmultiphase
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

Microstructure images of metals, alloys, and composites are usually characterized by humans or by supervised models that need expensive pixel-level labels. This paper claims that a largely automated, iterative three-step pipeline can do the job: first, an unsupervised score-vector method segments each micrograph into homogeneous regions; second, an uncertainty-aware classifier assigns each region to a known microstructure class or flags it as a new one; third, a supervised segmentation network refines the boundaries and, when applied repeatedly, the growing labeled database makes classification and segmentation more accurate. The authors demonstrate the pipeline on textured images and on real micrographs (silica-PMMA, dual-phase steel, Al-Zn, Pb-Sn, ultra-high carbon steel), reporting pixel-level segmentation accuracies from roughly 93 percent to over 99 percent on held-out test images. If the central claim is right, materials scientists can characterize massive high-throughput imaging data with minimal human inspection while new phases are discovered automatically rather than by memorizing every morphology.

What carries the argument

The load-bearing object is the Fisher score vector $s(\hat\theta; y_i, x_i)=\nabla_\theta \log P(X_i\mid N(X_i);\hat\theta)$ computed from a conditional pixel model. This vector has mean near zero when a pixel's local stochastic process matches the training distribution and nonzero mean when it does not, so spatially smoothed and clustered versions of these vectors define the unsupervised segmentation of Step 1. The Bayesian Gaussian mixture supplies a data-driven way to choose the number of homogeneous regions through posterior weights and AIC/BIC curves. Evidential deep learning replaces the classifier's softmax with a Dirichlet parameter output, producing an uncertainty $\hat u_i = K/\sum_k \alpha_{i,k}$ that flags regions not represented in the database. DeepLabv3+ with an Atrous Spatial Pyramid Pooling layer provides multi-scale pixel-wise segmentation, pretrained on synthetic collages of texture images. The loop's fourth component is the database itself: each iteration's verified labels are used as training data, and transfer learning lets the networks absorb a new class by adding one output channel.

What would settle it

Take a synthetic two-phase micrograph whose two phases are chosen so that their conditional pixel distributions match up to the neighborhood window used by the score model (for example, two textures that differ only through long-range correlations longer than the window), run Step 1, and check whether the Bayesian Gaussian mixture produces two clusters that match the painted regions; a single merge or split would show the separability assumption fails.

Watch

Extended reading notes

Core claim

The paper's central claim is that microstructure phase identification and segmentation should not be treated as a single supervised problem. Instead, the authors propose an iterative loop: fit one supervised learner to predict each pixel's gray level from its neighborhood, compute per-pixel Fisher score vectors, smooth and cluster them with Bayesian Gaussian mixtures to obtain homogeneous regions; classify each entire region with an AlexNet trained under evidential deep learning, using the network's uncertainty to decide whether the region belongs to a known class or a new one; and then train a DeepLabv3+ segmentation network on the resulting labeled regions, with synthetic collages from a texture database as pretraining. When new micrographs arrive, the loop repeats and the database grows. The paper reports that this loop identifies new classes reliably (held-out classes raise average EDL uncertainty from about 25 percent to about 61 percent) and that the supervised segmentation step achieves high pixel-level accuracy on several real materials data sets, with robustness to moderate magnification changes.

Load-bearing premise

The whole pipeline rests on the assumption that different phases produce separable Fisher score-vector clusters after a single model is fit across the image; if two phases have indistinguishable local stochastic signatures, the first segmentation step cannot separate them, and the later supervised steps inherit its mistakes.

Editorial extensions

If this is right

  • High-throughput imaging pipelines can replace per-image manual segmentation with a first-pass unsupervised pass, reducing the human effort from minutes per micrograph to reviewing only high-uncertainty regions.
  • The same framework can be applied to images from different instruments (SEM, TEM, X-ray CT, serial sectioning) without redesigning the pipeline.
  • New phases discovered in one batch are added to the database and immediately available for classifying later micrographs.
  • The supervised segmentation step can classify border pixels that the score-vector method cannot reach, and tolerates modest changes in magnification.
  • Transfer learning from non-material texture databases shortens training enough that the loop can be rerun as new classes arrive.

Reading between the lines

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

  • The paper does not specify how to set the EDL uncertainty threshold for declaring a new class; in practice that threshold trades off missed new phases against false alarms and would need calibration per application.
  • Because the score-vector method operates per pixel and the segmentation network is pixel-wise, the framework should extend to 3D tomographic volumes by treating slices or by using 3D neighborhoods, an extension the authors do not test.
  • The database-building loop suggests that classification accuracy should improve with each iteration, which could be measured directly by logging accuracy versus database size; this is a natural experimental check of the framework's central promise.
  • For microstructures with more geometric or periodic patterns, the authors themselves note the stochastic-texture assumption is a limitation, so the framework is best understood as covering statistically homogeneous random microstructures.
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

4 major / 6 minor

Summary. The paper proposes a three-step framework for microstructure image analysis. Step 1 uses Fisher score vectors from a fitted conditional pixel model, spatially smoothed and clustered with Bayesian Gaussian Mixture models, to segment multiphase micrographs into homogeneous regions in an unsupervised manner. Step 2 classifies each homogeneous region with an AlexNet-based evidential deep learning classifier, using the resulting uncertainty to flag regions that may belong to a new microstructure class. Step 3 trains a simplified DeepLabv3+ segmentation network using artificially augmented multiphase images and the database of labeled homogeneous regions built from Steps 1 and 2, with an iterative scheme that expands the database over time. The authors demonstrate each step on Brodatz textures, PMMA, dual-phase steel, Kylberg textures, Al-Zn, Pb-Sn, and UHCS datasets, reporting high accuracies for Steps 2 and 3 and a substantial separation in evidential uncertainty between known and held-out classes.

Significance. If validated as claimed, the framework would be a useful step toward automated, iterative microstructure characterization with minimal human labeling, addressing a real bottleneck in materials informatics. The paper's strengths include combining published score-vector methods with evidential deep learning and modern segmentation networks, and producing concrete quantitative results: 97-99% test accuracies on several segmentation benchmarks, and a clear aggregate separation in EDL uncertainty (24.58% vs. 61.25%) for held-out classes. The Brodatz-based iterative simulation in Section 3.4 also gives a reproducible demonstration of one aspect of the proposed database-building loop. The main weaknesses are that the foundational Step 1 is not validated on stationary single-phase controls, the full pipeline is not run end-to-end on real materials, and several load-bearing claims rely on heuristic or incomplete evaluations. The manuscript is a reasonable candidate for a methods journal if these validation gaps are addressed.

major comments (4)
  1. [Section 2.1, Eq. (2), Fig. 6] The zero-mean property of Fisher score vectors is derived under the assumption that the fitted conditional model is correctly specified within a stationary homogeneous region. The manuscript never demonstrates Step 1 on a stationary single-phase control; all Step 1 examples are deliberately nonstationary (PMMA agglomeration, dual-phase steel). Under model misspecification, local score-vector means can be nonzero even in a perfectly homogeneous micrograph, which would cause BGM to over-segment and create spurious HRs. Since Steps 2 and 3 inherit their training labels from Step 1, this is load-bearing for the entire iterative framework. The authors should add a stationary single-phase control experiment (e.g., a single Brodatz texture patch, and the spheroidite-only UHCS image in Fig. 1a) and report the BGM posterior weights and AIC/BIC elbow, showing that only one significant cluster emerges, or otherwise quantify the false-segmentation rate under misspecification.
  2. [Sections 3.3 and 3.4] The iterative database-building claim is demonstrated only on synthetic Brodatz collages. The real-material Step 3 experiments (Al-Zn, Pb-Sn, UHCS) use ground-truth pixel labels from the original datasets, not labels produced by Steps 1-2. The central claim that the framework builds a database from Steps 1 and 2 and uses it to improve segmentation on real multiphase materials is therefore not directly supported. To justify this claim, the manuscript should either run the full pipeline end-to-end on at least one real material dataset (with Step 1 HRs, Step 2 classification, and Step 3 training on the resulting labels), or explicitly re-scope the contribution as component validation plus a synthetic simulation of the iterative loop.
  3. [Section 3.3, Fig. 9] The paper asserts that Step 3 segmentation is more powerful than Step 1 segmentation, but no quantitative head-to-head comparison is made on the same real micrographs. Figure 6 shows Step 1 on PMMA and dual-phase steel, while Figure 9 shows Step 3 on Brodatz, Al-Zn, Pb-Sn, and UHCS; these are different datasets. The authors should report pixel-wise accuracy of Step 1 and Step 3 on identical test images, or otherwise provide a controlled comparison, to substantiate the claimed improvement.
  4. [Section 2.2 and Section 3.2] The EDL uncertainty mechanism is used to flag new microstructure classes, but the evaluation reports only aggregate means (24.58% for existing classes vs. 61.25% for held-out classes) for a single held-out split. No threshold selection procedure, receiver-operating-characteristic analysis, or false-positive/false-negative rates are given. Because 'new MC discovery' is a central contribution, the authors should define an operational decision rule (e.g., 'uncertainty above threshold T') and evaluate its detection and error rates, including how T is chosen.
minor comments (6)
  1. [Fig. 5 caption] The caption refers to the 'BMG method'; this should read 'BGM method' (the acronym used throughout the text).
  2. [Section 3.1, Fig. 1 caption] The Figure 1 caption contains stray text: 'Materials Microstructure Images0' and the trailing '0' in the caption; these appear to be artifacts and should be removed.
  3. [Section 3.3] The dataset sample sizes (e.g., 150,000 and 370,000 augmented images) are reported as 'sample sizes of augmented training, validation, and testing data splits'; it would be clearer to state explicitly that these are 256x256 patches generated from a smaller number of original micrographs.
  4. [Eq. (8)] The class-balancing weight is written as 'w0.5 j'; to avoid ambiguity, this should be typeset as w_j^{0.5} or described in text, since the fractional exponent is not immediately clear in the current notation.
  5. [General] The manuscript does not include a data/code availability statement. Given the emphasis on reproducibility and the use of public datasets (Brodatz, Kylberg, UHCS), the authors should add a statement about where the code and, where possible, processed data will be made available.
  6. [Section 3.2] The Kylberg classification experiments are on texture images, not materials micrographs. The authors should note this limitation more explicitly when drawing conclusions about the materials-oriented claims of Step 2, or add a small real-micrograph classification experiment for the classification network.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the framework's components are validated against external ground truth, and the self-cited score-vector method is independent support rather than a circular input.

full rationale

Step 1's score-vector segmentation is the only component that rests on the authors' prior work (Zhang et al. 2021; Zhang et al. 2023b). This is a self-citation, but it is independent support rather than circularity: the cited methods are externally peer-reviewed, Eq. (2) is a standard likelihood-score identity rather than a fitted relation, and the current paper revalidates Step 1 against external ground truth (dual-phase steel from Banerjee et al. 2013, Fig. 6b). Step 2's new-MC detection via EDL uncertainty is evaluated by holding out 4 of 28 Kylberg MCs and comparing average uncertainties (24.58% vs. 61.25%); no threshold was fit on the held-out classes and then reported as a prediction. Step 3's segmentation accuracy is computed against ground-truth masks for Brodatz, Al-Zn, Pb-Sn, and UHCS, with UHCS labels from DeCost et al. (2019), not from Steps 1-2. The iterative demonstration on Brodatz controls the MC count explicitly. The Conclusion's limitation (stochastic microstructures only) is a scope restriction, not a circularity admission; the 'results omitted for brevity' transfer-learning claim in Sec. 3.3 is an evidence gap, not a circular reduction. Overall, no equation sets a predicted quantity equal to a fitted or self-cited input by construction.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The framework does not introduce new physical entities. Its central load-bearing assumptions are the score-vector separability of microstructure classes and the correctness of the conditional pixel model, both inherited from prior work. The free parameters are engineering choices for windows, network size, and input resolution, plus a missing operational threshold for uncertainty-based new-class detection.

free parameters (5)
  • Neighborhood window half-length ls = 5
    Used in Step 1 experiments (Section 3.1) to define the local neighborhood for the conditional pixel model. Chosen by hand, not fit to data.
  • Spatial WMA window half-length lw = 20
    Used to smooth score vectors before clustering in Step 1 (Section 3.1). Chosen by hand.
  • Number of Xception layers in simplified DeepLabv3+ = 8
    Reduced from 20 to 8 to lower parameter count to about 1.4 million (Section 2.3). An architecture choice, not inferred from data.
  • Input image size for segmentation network = 256x256
    Reduced from the original 512x512 to fit the simplified network (Section 2.3).
  • EDL uncertainty threshold for flagging new MCs = Not specified
    The paper reports average uncertainty values for existing vs new classes (24.58% vs 61.25%) but does not give a threshold for declaring a new MC, which is required to operationalize Step 2.
assumptions (4)
  • domain assumption The conditional distribution P(Xi|N(Xi)) characterizes the joint distribution of all pixels via Gibbs sampling arguments.
    Stated in Section 2.1 as the basis for modeling the stochastic nature of the microstructure using a local conditional model.
  • domain assumption If the stochastic natures of different MCs are distinct, their Fisher score vectors center around different means and are separable.
    Section 2.1 states this as the key property enabling unsupervised segmentation. It is inherited from prior work (Zhang et al. 2021) and is not proven for all materials.
  • domain assumption The chosen supervised learning model (linear regression or neural network) is 'correct' enough to provide a useful fingerprint.
    The score vector analysis assumes the fitted model approximates the true conditional distribution; the paper notes robustness across model choices, but correctness of the model is assumed.
  • ad hoc to paper BGM with AIC/BIC or posterior weights correctly determines the number of homogeneous regions.
    Section 2.1 relies on these heuristics to select the cluster count; the paper demonstrates them on examples but does not establish general reliability.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Framework for Supervised and Unsupervised Segmentation and Classification of Materials Microstructure Images." pith.science (2026). https://pith.science/paper/CR7CD4SH

@misc{pith2026250207107,
  author       = {Pith},
  title        = {Pith review of: A Framework for Supervised and Unsupervised Segmentation and Classification of Materials Microstructure Images},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CR7CD4SH}},
  note         = {Machine review of arXiv:2502.07107}
}
abstract

Microstructure of materials is often characterized through image analysis to understand processing-structure-properties linkages. We propose a largely automated framework that integrates unsupervised and supervised learning methods to classify micrographs according to microstructure phase/class and, for multiphase microstructures, segments them into different homogeneous regions. With the advance of manufacturing and imaging techniques, the ultra-high resolution of imaging that reveals the complexity of microstructures and the rapidly increasing quantity of images (i.e., micrographs) enables and necessitates a more powerful and automated framework to extract materials characteristics and knowledge. The framework we propose can be used to gradually build a database of microstructure classes relevant to a particular process or group of materials, which can help in analyzing and discovering/identifying new materials. The framework has three steps: (1) segmentation of multiphase micrographs through a recently developed score-based method so that different microstructure homogeneous regions can be identified in an unsupervised manner; (2) {identification and classification of} homogeneous regions of micrographs through an uncertainty-aware supervised classification network trained using the segmented micrographs from Step $1$ with their identified labels verified via the built-in uncertainty quantification and minimal human inspection; (3) supervised segmentation (more powerful than the segmentation in Step $1$) of multiphase microstructures through a segmentation network trained with micrographs and the results from Steps $1$-$2$ using a form of data augmentation. This framework can iteratively characterize/segment new homogeneous or multiphase materials while expanding the database to enhance performance. The framework is demonstrated on various sets of materials and texture images.

Figures

Figures reproduced from arXiv: 2502.07107 by the authors.

Figure 1
Figure 1. Examples of single phase (a) and multiphase materials (b)-(f) [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Flow chart of the iterative framework. Between steps, the knowledge passed from [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Illustration of processing score vectors. (a) Neighborhood window (brown dashed [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Architectures of two Deep Learning neural networks used in Section 2.2 and [PITH_FULL_IMAGE:figures/full_fig_p018_4.png]
Figure 5
Figure 5. Figure 5: Results of the BMG method of Step 1 applied to the Brodatz data set. (a) A [PITH_FULL_IMAGE:figures/full_fig_p026_5.png]
Figure 6
Figure 6. Figure 6: The score-based nonstationarity analysis method of Step 1 applied to different [PITH_FULL_IMAGE:figures/full_fig_p028_6.png]
Figure 7
Figure 7. Figure 7: All micrographs in the Kylberg data set with their true label names on top of [PITH_FULL_IMAGE:figures/full_fig_p029_7.png]
Figure 8
Figure 8. Figure 8: Training and validation metrics for the Step 2 classification and ranking for the [PITH_FULL_IMAGE:figures/full_fig_p030_8.png]
Figure 9
Figure 9. Figure 9: Test results for the Step 3 supervised segmentation of the Brodatz and real ma [PITH_FULL_IMAGE:figures/full_fig_p031_9.png]
Figure 10
Figure 10. Figure 10: Training (tr) and validation (val) accuracy for two ways of training when iter [PITH_FULL_IMAGE:figures/full_fig_p036_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

12 extracted references · 10 canonical work pages

  1. [1]

    AFRL (2018) Air Force Research Laboratory (AFRL) additive manufac- turing (AM) modeling challenge series.,

    AFRL (2018), “AFRL (2018) Air Force Research Laboratory (AFRL) additive manufac- turing (AM) modeling challenge series.,”. Agrawal, Ankit and Choudhary, Alok (2016), “Perspective: Materials informatics and big data: Realization of the “fourth paradigm” of science in materials science,” Apl Materials, 4, 5, 053208. — (2019), “Deep materials informatics: Ap...

    work page 2018

  2. [5]

    Extraction of reduced-order process-structure linkages from phase-field simulations,

    Yabansu, Yuksel C, Steinmetz, Philipp, H¨ otzer, Johannes, Kalidindi, Surya R, and Nestler, Britta (2017), “Extraction of reduced-order process-structure linkages from phase-field simulations,” Acta Materialia , 124, 182–194. You, Jane and Cohen, Harvey A (1993), “Classification and segmentation of rotated and scaled textured images using texture “tuned” ...

    work page 2017

  3. [7]

    Unleashing the power of self-supervised image denoising: A comprehen- sive reviewa,

    Zhang, Dan, Zhou, Fangfang, Albu, Felix, Wei, Yuanzhou, Yang, Xiao, Gu, Yuan, and Li, Qiang (2023a), “Unleashing the power of self-supervised image denoising: A comprehen- sive reviewa,” arXiv preprint arXiv:2308.00247 a. Zhang, Kungang, Apley, Daniel W, and Chen, Wei (2021), “Nonstationarity Analysis of Materials Microstructures via Fisher Score Vectors,...

    arXiv 2023

  4. [15]

    Filtering for texture classification: A comparative study,

    Randen, Trygve and Husoy, John Hakon (1999), “Filtering for texture classification: A comparative study,” IEEE Transactions on pattern analysis and machine intelligence , 21, 4, 291–310. Ronneberger, Olaf, Fischer, Philipp, and Brox, Thomas (2015),, U-net: Convolutional networks for biomedical image segmentation. in International Conference on Medical ima...

    work page 1999

  5. [17]

    XSEDE: accelerating scientific discovery,

    Tivive, Fok Hing Chi and Bouzerdoum, Abdesselam (2006),, Texture classification using convolutional neural networks. in TENCON 2006-2006 IEEE Region 10 Conference , 1–4. IEEE. Todorovic, Sinisa and Ahuja, Narendra (2009),, Texel-based texture segmentation. in 2009 IEEE 12th International Conference on Computer Vision , 841–848. IEEE. Towns, John, Cockeril...

    work page 2006

  6. [31]

    Optimizing convo- lutional neural networks to perform semantic segmentation on large materials imaging datasets: X-ray tomography and serial sectioning,

    47 Stan, Tiberiu, Thompson, Zachary T, and Voorhees, Peter W (2020), “Optimizing convo- lutional neural networks to perform semantic segmentation on large materials imaging datasets: X-ray tomography and serial sectioning,” Materials Characterization , 110119. Strohmann, Tobias, Bugelnig, Katrin, Breitbarth, Eric, Wilde, Fabian, Steffens, Thomas, Germann,...

    work page 2020

  7. [74]

    Incremental learning to segment micrographs,

    Impoco, Gaetano and Tuminello, L (2015), “Incremental learning to segment micrographs,” Computer Vision and Image Understanding , 140, 144–152. Jain, Anubhav, Hautier, Geoffroy, Ong, Shyue Ping, and Persson, Kristin (2016a), “New opportunities for materials informatics: resources and data mining techniques for uncov- ering hidden relationshipsa,” Journal ...

    work page 2015

  8. [258]

    Characterization and design of functional quasi- random nanostructured materials using spectral density function,

    Yu, Shuangcheng, Zhang, Yichi, Wang, Chen, Lee, Won-kyu, Dong, Biqin, Odom, Teri W, Sun, Cheng, and Chen, Wei (2017), “Characterization and design of functional quasi- random nanostructured materials using spectral density function,”Journal of Mechanical Design, 139,

    work page 2017

Show all 12 references
  1. [370]

    Imagenet large scale visual recognition challenge,

    Russakovsky, Olga, Deng, Jia, Su, Hao, Krause, Jonathan, Satheesh, Sanjeev, Ma, Sean, Huang, Zhiheng, Karpathy, Andrej, Khosla, Aditya, Bernstein, Michael, et al. (2015), “Imagenet large scale visual recognition challenge,” International journal of computer vision, 115, 3, 211...

    2015

  2. [617]

    Additive manufacturing of ceramic compo- nents,

    Moritz, Tassilo and Maleksaeedi, Saeed (2018), “Additive manufacturing of ceramic compo- nents,” In Additive Manufacturing, ed. I. Additive Manufacturing, Elsevier, pp. 105–161. O’Mara, Jordan, Meredig, Bryce, and Michel, Kyle (2016), “Materials data infrastructure: a case stu...

    2018

  3. [1258]

    Fundamentals, processes and applications of high-permittivity polymer–matrix composites,

    Dang, Zhi-Min, Yuan, Jin-Kai, Zha, Jun-Wei, Zhou, Tao, Li, Sheng-Tao, and Hu, Guo-Hua (2012), “Fundamentals, processes and applications of high-permittivity polymer–matrix composites,” Progress in materials science , 57, 4, 660–723. DeCost, Brian L, Francis, Toby, and Holm, El...

    2012

  4. [3900]

    From organized high-throughput data to phenomenological theory using machine learning: the example of dielectric breakdown,

    Karoui, Imen, Fablet, Ronan, Boucher, J-M, and Augustin, J-M (2006),, Region-based image segmentation using texture statistics and level-set methods. in 2006 IEEE Inter- national Conference on Acoustics Speech and Signal Processing Proceedings, vol. 2, II–II. IEEE. Kim, Chiho,...

    2006 arXiv

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.