Pith. sign in

REVIEW 3 major objections 5 minor 2 cited by

ShapeEmbed: a self-supervised learning framework for 2D contour quantification

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a self-supervised variational autoencoder trained on Euclidean distance matrices of 2D contours produces shape descriptors that are invariant to translation, scaling, rotation, reflection, and contour point…

desk verdict ShapeEmbed is a solid, well-evaluated method for learning shape descriptors from distance matrices, but the paper's central guarantee of point-indexation invariance is overstated and needs a direct test before I'd trust it. read the letter →

arxiv 2507.01009 v1 pith:LIEQO6QS submitted 2025-07-01 cs.CV q-bio.QM

Anna Foix Romero , Craig Russell , Alexander Krull , Virginie Uhlmann This is my paper · ORCID
classification cs.CVq-bio.QM
keywords shapedescriptorsvariationalautoencoderdistancematrixgeometricinvarianceself-supervisedlearningcontourquantificationbiologicalimagingrepresentation
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

ShapeEmbed aims to show that a self-supervised variational autoencoder can turn the outline of any simply connected 2D object into a fixed-length descriptor that does not change when the object is translated, scaled, rotated, mirrored, or when the contour's starting point and travel direction change. The paper's evidence is a set of classification experiments in which a logistic regression on ShapeEmbed's latent codes beats classical shape statistics, Fourier descriptors, contrastive learning, masked autoencoders, and the orientation-invariant O2VAE on MNIST, MPEG-7, and two biological datasets. The practical payoff would be a shape representation that needs no pre-alignment or normalization of objects before analysis, which matters for microscopy where cells and organisms appear at arbitrary pose and size.

What carries the argument

The load-bearing object is the Euclidean distance matrix of the contour points: an N by N array of pairwise point distances that is translation- and rotation-invariant by construction and scale-invariant after division by its Frobenius norm. The argument is carried by three components: circular padding in every convolution and pooling layer of a ResNet-18 encoder to make it shift-equivariant under diagonal re-indexing; summing the encoder outputs of the matrix and its horizontal and vertical mirror to add reflection invariance; and a reconstruction loss that takes the minimum over all 2N equivalent indexations so the decoder is never forced to commit to one indexing.

What would settle it

Take a single contour, compute its distance matrix, generate all 2N re-indexed versions (N starting points times two travel directions), push each through the trained encoder, and measure the maximum pairwise distance between the resulting latent codes. If any two re-indexings yield codes that differ by more than the typical within-class spread of the latent space, the guaranteed-invariance claim is false, and the paper reports no such direct test.

Watch

Extended reading notes

Core claim

The central claim is that encoding the Euclidean distance matrix of a contour, rather than the segmentation mask, with a VAE whose encoder uses circular padding and whose decoder is trained with a min-over-reindexations loss yields latent vectors that are invariant to similarity transformations and to contour re-indexing. The paper argues that distance matrices are inherently invariant to translation and rotation, become scale-invariant after Frobenius normalization, and that the remaining indexation ambiguity, the 2N equivalent matrices from choice of origin and travel direction, is removed by the encoder design and loss, so the same latent code is produced for every equivalent indexing. It further reports that this descriptor outperforms the tested baselines in shape classification on natural image benchmarks and biological imaging datasets, and that adding object size back as an optional feature improves results on biology tasks.

Load-bearing premise

The whole invariance story depends on the claim that replacing padding with circular padding makes the ResNet-18 encoder shift-equivariant enough that all 2N re-indexed distance matrices of a contour map to the same latent code, even though the paper admits that stride and pooling break true shift equivariance.

Editorial extensions

If this is right

  • Shape descriptors can be extracted without alignment, centering, or size normalization of input objects, removing a preprocessing step that classical shape pipelines require.
  • The same trained model can reconstruct outlines from latent codes via multidimensional scaling, enabling visualization and generative sampling of shapes directly from the descriptor space.
  • Reintroducing object size as an extra feature alongside the invariant code improves classification on biological datasets, so size can be treated as an optional, task-dependent channel.
  • The method applies to simply connected 2D contours and would need new machinery for contours with holes or for genuine 3D surfaces.
  • In biological imaging, the unsupervised latent structure can reveal mismatches between experimental labels and actual phenotypes, as demonstrated on the C. elegans dataset.

Reading between the lines

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

  • If the invariance claim holds exactly, ShapeEmbed could act as a drop-in shape encoder for unsupervised phenotype discovery, where structure in the latent space is inspected directly rather than through a trained classifier.
  • The min-over-reindexations loss may let the decoder absorb residual encoder non-invariance rather than enforcing exact latent equality, so an explicit equality test would be needed to separate the two effects.
  • The consistent gains from adding size back suggest the most useful practical descriptor is a two-part code: one scale-invariant shape vector plus one scalar scale channel, rather than a single invariant vector.
  • The same distance-matrix-plus-indexation-invariance recipe could transfer to other self-supervised architectures than VAEs, provided their encoders can be made shift-equivariant under the 2N re-indexing group.
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 introduces ShapeEmbed, a self-supervised variational autoencoder that encodes ordered 2D contour points through normalized Euclidean distance matrices into a latent shape descriptor. The authors claim invariance to translation, scaling, rotation, reflection, and contour point re-indexing. The encoder is a ResNet-18 modified with circular padding, and the decoder is trained with a reconstruction loss that minimizes over all 2N equivalent re-indexings of the input distance matrix. The method is evaluated on MNIST, MPEG-7, BBBC010, and MEF datasets against classical descriptors, SimCLR, MAE variants, and O2VAE, with additional ablations and a robustness study on degraded segmentations.

Significance. If the invariance claims hold, ShapeEmbed would be a practically valuable contribution: it would provide the first VAE-based shape descriptor that is fully invariant to similarity transformations and to contour re-indexing, with demonstrated strengths in downstream classification and in generative exploration of bioimaging shape spaces. The empirical work is extensive and carefully reported: multiple baselines, multiple metrics with 5-fold cross-validation, separate ablations for rotation/translation and for regularization terms, a segmentation-noise robustness study, and qualitative latent-space analyses. The central limitation is that the paper's strongest advertised property, exact indexation invariance, is asserted but not directly verified, and the architecture's own text acknowledges that exact shift equivariance is lost in strided and pooling layers.

major comments (3)
  1. [Section 3.2] The sentence 'our VAE is thus guaranteed to map all possible equivalent indexings of successive outline points to the same latent vector' is stronger than what the architecture supports. The authors themselves state that, because ResNet-18 uses stride and pooling, 'the result is no longer truly shift equivariant or invariant' even with circular padding. Circular padding in convolutional layers does not restore exact equivariance through strided convolutions and pooling. No experiment in Section 4 measures latent vectors for the 2N re-indexed distance matrices of a single contour. This is a load-bearing assertion for contribution 1, so the manuscript should either provide a direct invariance test (for example, maximum pairwise latent distance or cosine similarity across D^{k,o} for held-out contours) or explicitly weaken the claim to approximate/practical invariance.
  2. [Equation (2)] The min-over-reindexations reconstruction loss supervises only the decoder: it selects the version of the ground-truth matrix closest to the decoder output. This loss is minimized even if the encoder preserves indexation information, because the decoder can align to whichever indexation was preserved. Nothing in the objective directly enforces equality of latent codes across the 2N equivalent matrices. The claim in Section 3.2 that the encoder 'is unable to distinguish' between re-indexings is therefore not implied by the loss alone. The authors should add an explicit invariance term or, at minimum, an empirical measurement showing that encoder outputs for D and D^{k,o} coincide.
  3. [Table 2] The 'No index. inv.' ablation removes circular padding and the custom indexation-invariant loss simultaneously, replacing both with constant padding and standard MSE reconstruction. Consequently, any observed drop in F1-score cannot be attributed to either component individually. Because this is the only experiment directly tied to indexation invariance, the ablation should separate the two changes. Moreover, downstream logistic-regression accuracy is a coarse proxy that tolerates residual intra-class latent variation; a direct latent-equality metric is needed to support the invariance claim.
minor comments (5)
  1. [Table 5] The rows for MEF and BBBC010 appear to be swapped relative to Section 4.1: Section 4.1 reports 1,407 objects for BBBC010 and 26,198 objects for MEF, while Table 5 assigns 1,407 to MEF and 26,198 to BBBC010. The maximum/minimum outline sizes appear swapped as well.
  2. [Section 4 / Supplementary B] The code repository is given as 'https://github.com/link_to_be_added_in_camera-ready_version', which is a placeholder. Since reproducibility is a stated goal, a working repository link or an explicit statement of availability is needed.
  3. [Section 4.5] The 'ShapeEmbed+Sz' variant is used in Table 4 but not defined in the main text. Please specify whether the saved distance-matrix norm is concatenated to the latent vector before the logistic regression classifier and how that feature is normalized.
  4. [Supplementary Section G.3] The sentence 'ShapeEmbed performs well (F1-score≤ 0.7)' appears to be a typo: the reported values and Figure 12 indicate that the intended statement is F1-score ≥ 0.7.
  5. [Throughout] There are small typographical and citation issues, including 'not to to the same solution' in Section 3.4 and the spelling 'Lorense' for the marching-squares reference, which should be 'Lorensen'.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: ShapeEmbed's invariance construction and benchmark evaluations are self-contained; the untested indexation-invariance guarantee is a soundness concern, not circular reasoning.

full rationale

The paper's claimed derivation chain is self-contained rather than circular. Translation and rotation invariance follow from the Euclidean distance matrix definition with explicit algebra in Supplementary Section A. Scale invariance follows from the Frobenius-norm scaling identity (Eqs. 4-6). Re-indexed matrices are defined by Eq. 1, and the encoder modifications (circular padding and summed mirrored passes) are architectural constructions. The custom reconstruction loss (Eq. 2) is a training objective that minimizes over re-indexed versions of the input; it is not a fitted parameter renamed as a prediction, and the downstream classification evaluations use external labels on MNIST, MPEG-7, BBBC010, and MEF. The only self-citation in the introduction (Zinchenko et al.) is motivational and not load-bearing. The paper itself admits that with stride and pooling 'the result is no longer truly shift equivariant or invariant' (Section 3.2), so the assertion that the VAE is 'guaranteed to map all possible equivalent indexings of successive outline points to the same latent vector' is stronger than what the architecture provably delivers, and no experiment directly measures latent-code equality. This is an unverified assumption and a correctness risk, not circularity, because the claim does not reduce to its inputs by definition and the method's empirical value is assessed against independent baselines.

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

The central claim rests on standard distance-matrix geometry plus two unverified assumptions about the encoder: that circular padding yields sufficient shift equivariance despite stride and pooling, and that the min-loss does not mask retained indexation information. Hyperparameters are set by hand or tuning; no new physical entities are introduced.

free parameters (4)
  • N (contour point count) = 64
    Number of uniformly sampled outline points; chosen by hand and used for all datasets; affects representation fidelity and the 2N re-indexations in the loss.
  • beta (KL weight) = 1e-10
    Weight on the KL divergence in Eq. (3); set empirically and stated to balance reconstruction and sampling.
  • gamma, delta, epsilon (regularization weights) = 1e-5 each
    Weights on diagonal, non-negativity, and symmetry losses in Eq. (3); found through hyperparameter tuning.
  • Latent dimension = 128
    Dimensionality of the shape descriptor; fixed in all experiments.
assumptions (4)
  • domain assumption Contours are simply connected and can be represented by an ordered sequence of points
    Section 3.1 states ShapeEmbed operates with simply connected contours; holes or multiple components are excluded, which limits biological applicability.
  • domain assumption The distance matrix, after normalization, uniquely describes the contour shape up to isometry and indexation
    Section 3.1 relies on MDS to reconstruct outlines from distance matrices; this is a standard property of Euclidean distance matrices.
  • ad hoc to paper Circular padding makes the modified ResNet-18 sufficiently shift equivariant for exact indexation invariance
    Section 3.2 states the encoder is 'guaranteed' to map re-indexings to the same latent vector, but immediately concedes stride and pooling break true shift equivariance and only 'in practice' the effect is observed.
  • ad hoc to paper The min-over-reindexations reconstruction loss (Eq. 2) does not allow the encoder to retain indexation information
    Because the loss is minimized over all 2N indexed versions of the input, a non-invariant encoder could still achieve low reconstruction error by keeping indexation in the latent code; no experiment verifies latent invariance directly.

how reviews work

0 comments
Cite this review

Pith. "Pith review of ShapeEmbed: a self-supervised learning framework for 2D contour quantification." pith.science (2026). https://pith.science/paper/LIEQO6QS

@misc{pith2026250701009,
  author       = {Pith},
  title        = {Pith review of: ShapeEmbed: a self-supervised learning framework for 2D contour quantification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LIEQO6QS}},
  note         = {Machine review of arXiv:2507.01009}
}
read the original abstract

The shape of objects is an important source of visual information in a wide range of applications. One of the core challenges of shape quantification is to ensure that the extracted measurements remain invariant to transformations that preserve an object's intrinsic geometry, such as changing its size, orientation, and position in the image. In this work, we introduce ShapeEmbed, a self-supervised representation learning framework designed to encode the contour of objects in 2D images, represented as a Euclidean distance matrix, into a shape descriptor that is invariant to translation, scaling, rotation, reflection, and point indexing. Our approach overcomes the limitations of traditional shape descriptors while improving upon existing state-of-the-art autoencoder-based approaches. We demonstrate that the descriptors learned by our framework outperform their competitors in shape classification tasks on natural and biological images. We envision our approach to be of particular relevance to biological imaging applications.

Figures

Figures reproduced from arXiv: 2507.01009 by the authors.

Figure 1
Figure 1. Overview of ShapeEmbed. ShapeEmbed converts the outline of an object from a 2D [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Effect of indexation changes on the distance matrix. A point outline (a) and its cor [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Sample images from the BBBC010 dataset illustrating the (a) live and (c) dead experimental [PITH_FULL_IMAGE:figures/full_fig_p016_3.png] view at source ↗
Figures from the paper (11 more)
Figure 4
Figure 4. Figure 4: Sample images from the MEF dataset illustrating the (a) control (non-patterned), (c) [PITH_FULL_IMAGE:figures/full_fig_p017_4.png]
Figure 5
Figure 5. Figure 5: Projection (t-SNE) of the rMNIST latent space. (a) The latent representation of rMNIST [PITH_FULL_IMAGE:figures/full_fig_p020_5.png]
Figure 6
Figure 6. Figure 6: Effect of distance matrix regularization on outline reconstruction. For a randomly-picked [PITH_FULL_IMAGE:figures/full_fig_p022_6.png]
Figure 7
Figure 7. Figure 7: Mean outline generation in the MPEG-7 dataset. We illustrate [PITH_FULL_IMAGE:figures/full_fig_p023_7.png]
Figure 8
Figure 8. Figure 8: Novel outline generation in the MNIST dataset. We illustrate [PITH_FULL_IMAGE:figures/full_fig_p023_8.png]
Figure 9
Figure 9. Figure 9: Projection (t-SNE) of the BBBC010 latent space learned by ShapeEmbed. Data points [PITH_FULL_IMAGE:figures/full_fig_p024_9.png]
Figure 10
Figure 10. Figure 10: Projection (t-SNE) of the MEF latent space learned by ShapeEmbed. Data points are [PITH_FULL_IMAGE:figures/full_fig_p025_10.png]
Figure 11
Figure 11. Figure 11: Sample images from the "circle" class of the MEF dataset illustrating the various noise [PITH_FULL_IMAGE:figures/full_fig_p026_11.png]
Figure 12
Figure 12. Figure 12: F1-score across varying levels of noise levels degrading the MEF dataset. The x-axis [PITH_FULL_IMAGE:figures/full_fig_p027_12.png]
Figure 13
Figure 13. Figure 13: Mean outline generation in the BBBC010 dataset. We illustrate [PITH_FULL_IMAGE:figures/full_fig_p028_13.png]
Figure 14
Figure 14. Figure 14: Novel outline generation in the BBBC010 dataset. We illustrate [PITH_FULL_IMAGE:figures/full_fig_p028_14.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

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

  1. The Push-Forward Transform for Continuous and Robust Comparison of Dynamic Shapes

    cs.CV 2026-08 conditional novelty 5.0 of 10

    Mapping signed distance functions through a diffeomorphism to a common reference domain yields a spectral shape descriptor (PF-SDM) that is invariant to rigid motions and scaling, and is competitive with deep learning...

  2. Attention Mechanism in Randomized Time Warping

    cs.CV 2025-08 reject novelty 5.0 of 10

    The abstract claims RTW's learned weights are interpretable as self-attention with a 5% benchmark gain, but the submitted full text is an unrelated contour-learning paper, so the claims are unsupported.

Reference graph

Works this paper leans on

62 extracted references · 43 canonical work pages · cited by 2 Pith papers

  1. [1]

    Identification of everyday objects on the basis of silhouette and outline versions

    Johan Wagemans, Joeri De Winter, Hans Op de Beeck, Annemie Ploeger, Tom Beckers, and Peter Vanroose. Identification of everyday objects on the basis of silhouette and outline versions. Perception, 37 0 (2): 0 207--244, 2008

    work page 2008

  2. [2]

    Statistical shape analysis: with applications in R, volume 995

    Ian L Dryden and Kanti V Mardia. Statistical shape analysis: with applications in R, volume 995. John Wiley & Sons, 2016

    work page 2016

  3. [3]

    Biology and physics of cell shape changes in development

    Ewa Paluch and Carl-Philipp Heisenberg. Biology and physics of cell shape changes in development. Current Biology, 19 0 (17): 0 R790--R799, 2009

    work page 2009

  4. [4]

    Decoding information in cell shape

    Padmini Rangamani, Azi Lipshtat, Evren U Azeloglu, Rhodora Cristina Calizo, Mufeng Hu, Saba Ghassemi, James Hone, Suzanne Scarlata, Susana R Neves, and Ravi Iyengar. Decoding information in cell shape. Cell, 154 0 (6): 0 1356--1369, 2013

    work page 2013

  5. [5]

    Cell and nucleus shape as an indicator of tissue fluidity in carcinoma

    Steffen Grosser, J \"u rgen Lippoldt, Linda Oswald, Matthias Merkel, Daniel M Sussman, Fr \'e d \'e ric Renner, Pablo Gottheil, Erik W Morawetz, Thomas Fuhs, Xiaofan Xie, et al. Cell and nucleus shape as an indicator of tissue fluidity in carcinoma. Physical Review X, 11 0 (1): 0 011033, 2021

    work page 2021

  6. [6]

    Morphofeatures for unsupervised exploration of cell types, tissues, and organs in volume electron microscopy

    Valentyna Zinchenko, Johannes Hugger, Virginie Uhlmann, Detlev Arendt, and Anna Kreshuk. Morphofeatures for unsupervised exploration of cell types, tissues, and organs in volume electron microscopy. Elife, 12: 0 e80918, 2023

    work page 2023

  7. [7]

    Image-based multivariate profiling of drug responses from single cells

    Lit-Hsin Loo, Lani F Wu, and Steven J Altschuler. Image-based multivariate profiling of drug responses from single cells. Nature methods, 4 0 (5): 0 445--453, 2007

    work page 2007

  8. [8]

    Visualizing cellular imaging data using phenoplot

    Heba Z Sailem, Julia E Sero, and Chris Bakal. Visualizing cellular imaging data using phenoplot. Nature communications, 6 0 (1): 0 5825, 2015

    work page 2015

Show all 62 references
  1. [9]

    Comparison of quantitative methods for cell-shape analysis

    Zachary Pincus and JA Theriot. Comparison of quantitative methods for cell-shape analysis. Journal of microscopy, 227 0 (2): 0 140--156, 2007

    2007

  2. [10]

    Reducing the dimensionality of data with neural networks

    Geoffrey E Hinton and Ruslan R Salakhutdinov. Reducing the dimensionality of data with neural networks. science, 313 0 (5786): 0 504--507, 2006

    2006

  3. [11]

    Kingma and Max Welling

    Diederik P. Kingma and Max Welling. Auto-encoding variational bayes. In Proceedings of the 2nd International Conference on Learning Representations (ICLR), Banff, AB, Canada, 2014. URL https://arxiv.org/abs/1312.6114

    2014 arXiv

  4. [12]

    Quantitative comparison of principal component analysis and unsupervised deep learning using variational autoencoders for shape analysis of motile cells

    Caleb K Chan, Amalia Hadjitheodorou, Tony Y-C Tsai, and Julie A Theriot. Quantitative comparison of principal component analysis and unsupervised deep learning using variational autoencoders for shape analysis of motile cells. bioRxiv, pages 2020--06, 2020

    2020

  5. [13]

    Evaluation of methods for generative modeling of cell and nuclear shape

    Xiongtao Ruan and Robert F Murphy. Evaluation of methods for generative modeling of cell and nuclear shape. Bioinformatics, 35 0 (14): 0 2475--2485, 2019

    2019

  6. [14]

    Kendall shape-vae: Learning shapes in a generative framework

    Sharvaree Vadgama, Jakub Mikolaj Tomczak, and Erik J Bekkers. Kendall shape-vae: Learning shapes in a generative framework. In NeurIPS 2022 Workshop on Symmetry and Geometry in Neural Representations, New Orleans, LA, USA, 2022

    2022

  7. [15]

    Continuous kendall shape variational autoencoders

    Sharvaree Vadgama, Jakub M Tomczak, and Erik Bekkers. Continuous kendall shape variational autoencoders. In International Conference on Geometric Science of Information, pages 73--81, Saint-Malo, France, 2023. Springer

    2023

  8. [16]

    Euclidean distance matrices: essential theory, algorithms, and applications

    Ivan Dokmanic, Reza Parhizkar, Juri Ranieri, and Martin Vetterli. Euclidean distance matrices: essential theory, algorithms, and applications. IEEE Signal Processing Magazine, 32 0 (6): 0 12--30, 2015

    2015

  9. [17]

    Multidimensional scaling

    Trevor F Cox and Michael AA Cox. Multidimensional scaling. CRC press, 2000

    2000

  10. [18]

    Multiscale distance matrix for fast plant leaf recognition

    Rongxiang Hu, Wei Jia, Haibin Ling, and Deshuang Huang. Multiscale distance matrix for fast plant leaf recognition. IEEE transactions on image processing, 21 0 (11): 0 4667--4672, 2012

    2012

  11. [19]

    Wesd--weighted spectral distance for measuring shape dissimilarity

    Ender Konukoglu, Ben Glocker, Antonio Criminisi, and Kilian M Pohl. Wesd--weighted spectral distance for measuring shape dissimilarity. IEEE transactions on pattern analysis and machine intelligence, 35 0 (9): 0 2284--2297, 2012

    2012

  12. [20]

    Cajal enables analysis and integration of single-cell morphological data using metric geometry

    Kiya W Govek, Patrick Nicodemus, Yuxuan Lin, Jake Crawford, Artur B Saturnino, Hannah Cui, Kristi Zoga, Michael P Hart, and Pablo G Camara. Cajal enables analysis and integration of single-cell morphological data using metric geometry. Nature Communications, 14 0 (1): 0 3672, 2023

    2023

  13. [21]

    Highly accurate protein structure prediction with alphafold

    John Jumper, Richard Evans, Alexander Pritzel, Tim Green, Michael Figurnov, Olaf Ronneberger, Kathryn Tunyasuvunakool, Russ Bates, Augustin Z \' dek, Anna Potapenko, et al. Highly accurate protein structure prediction with alphafold. nature, 596 0 (7873): 0 583--589, 2021

    2021

  14. [22]

    S ch\"onberger, J uan Nunez-Iglesias , F ran c ois B oulogne, J oshua D

    S t\'efan van der Walt, J ohannes L . S ch\"onberger, J uan Nunez-Iglesias , F ran c ois B oulogne, J oshua D . W arner, N eil Y ager, E mmanuelle G ouillart, T ony Y u, and the scikit-image contributors. scikit-image: image processing in P ython. PeerJ, 2: 0 e453, 6 2014. doi...

    2014 doi

  15. [23]

    Quantitative morphological signatures define local signaling networks regulating cell morphology

    Chris Bakal, John Aach, George Church, and Norbert Perrimon. Quantitative morphological signatures define local signaling networks regulating cell morphology. science, 316 0 (5832): 0 1753--1756, 2007

    2007

  16. [24]

    Identification of phenotype-specific networks from paired gene expression--cell shape imaging data

    Charlie George Barker, Eirini Petsalaki, Girolamo Giudice, Julia Sero, Emmanuel Nsa Ekpenyong, Chris Bakal, and Evangelia Petsalaki. Identification of phenotype-specific networks from paired gene expression--cell shape imaging data. Genome Research, 32 0 (4): 0 750--765, 2022

    2022

  17. [25]

    Shape discrimination using fourier descriptors

    Eric Persoon and King-Sun Fu. Shape discrimination using fourier descriptors. IEEE Transactions on systems, man, and cybernetics, 7 0 (3): 0 170--179, 1977

    1977

  18. [26]

    Elliptic fourier features of a closed contour

    Frank P Kuhl and Charles R Giardina. Elliptic fourier features of a closed contour. Computer Graphics and Image Processing, 18 0 (3): 0 236--258, 1982. doi:https://doi.org/10.1016/0146-664X(82)90034-X

    1982 doi

  19. [27]

    A simple framework for contrastive learning of visual representations

    Ting Chen, Simon Kornblith, Mohammad Norouzi, and Geoffrey Hinton. A simple framework for contrastive learning of visual representations. In International conference on machine learning, pages 1597--1607. PmLR, 2020

    2020

  20. [28]

    Momentum contrast for unsupervised visual representation learning

    Kaiming He, Haoqi Fan, Yuxin Wu, Saining Xie, and Ross Girshick. Momentum contrast for unsupervised visual representation learning. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 9729--9738, 2020

    2020

  21. [29]

    Emerging properties in self-supervised vision transformers

    Mathilde Caron, Hugo Touvron, Ishan Misra, Herv \'e J \'e gou, Julien Mairal, Piotr Bojanowski, and Armand Joulin. Emerging properties in self-supervised vision transformers. In Proceedings of the IEEE/CVF international conference on computer vision, pages 9650--9660, 2021

    2021

  22. [30]

    Masked autoencoders are scalable vision learners

    Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Doll \'a r, and Ross Girshick. Masked autoencoders are scalable vision learners. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 16000--16009, 2022

    2022

  23. [31]

    Jamieson, Erik S

    Assaf Zaritsky, Andrew R. Jamieson, Erik S. Welf, Andres Nevarez, Justin Cillay, Ugur Eskiocak, Brandi L. Cantarel, and Gaudenz Danuser. Interpretable deep learning uncovers cellular properties in label-free live cell images that are predictive of highly metastatic melanoma. C...

    2021

  24. [32]

    Orientation-invariant autoencoders learn robust representations for shape profiling of cells and organelles

    James Burgess, Jeffrey J Nirschl, Maria-Clara Zanellati, Alejandro Lozano, Sarah Cohen, and Serena Yeung-Levy. Orientation-invariant autoencoders learn robust representations for shape profiling of cells and organelles. Nature Communications, 15 0 (1): 0 1022, 2024. doi:10.103...

    2024 doi

  25. [33]

    Invariant shape representation learning for image classification

    Tonmoy Hossain, Jing Ma, Jundong Li, and Miaomiao Zhang. Invariant shape representation learning for image classification. arXiv preprint arXiv:2411.12201, 2024

    2024 arXiv

  26. [34]

    A rotation-invariant framework for deep point cloud analysis

    Xianzhi Li, Ruihui Li, Guangyong Chen, Chi-Wing Fu, Daniel Cohen-Or, and Pheng-Ann Heng. A rotation-invariant framework for deep point cloud analysis. IEEE transactions on visualization and computer graphics, 28 0 (12): 0 4503--4514, 2021 a

    2021

  27. [35]

    Riconv++: Effective rotation invariant convolutions for 3d point clouds deep learning

    Zhiyuan Zhang, Binh-Son Hua, and Sai-Kit Yeung. Riconv++: Effective rotation invariant convolutions for 3d point clouds deep learning. International Journal of Computer Vision, 130 0 (5): 0 1228--1243, 2022

    2022

  28. [36]

    A closer look at rotation-invariant deep point cloud analysis

    Feiran Li, Kent Fujiwara, Fumio Okura, and Yasuyuki Matsushita. A closer look at rotation-invariant deep point cloud analysis. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 16218--16227, 2021 b

    2021

  29. [37]

    Ri-mae: Rotation-invariant masked autoencoders for self-supervised point cloud representation learning

    Kunming Su, Qiuxia Wu, Panpan Cai, Xiaogang Zhu, Xuequan Lu, Zhiyong Wang, and Kun Hu. Ri-mae: Rotation-invariant masked autoencoders for self-supervised point cloud representation learning. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 39, pages 701...

    2025

  30. [38]

    Self-supervised learning of rotation-invariant 3d point set features using transformer and its self-distillation

    Takahiko Furuya, Zhoujie Chen, Ryutarou Ohbuchi, and Zhenzhong Kuang. Self-supervised learning of rotation-invariant 3d point set features using transformer and its self-distillation. Computer Vision and Image Understanding, 244: 0 104025, 2024

    2024

  31. [39]

    General e(2) -equivariant steerable cnns

    Maurice Weiler and Gabriele Cesa. General e(2) -equivariant steerable cnns. In Advances in Neural Information Processing Systems (NeurIPS), volume 32, pages 5214--5224, Vancouver, BC, Canada, 2019. URL https://arxiv.org/abs/1911.08251

    2019 arXiv

  32. [40]

    A high resolution 3d surface construction algorithm

    WE Lorense. A high resolution 3d surface construction algorithm. Proc. Siggraph 87, 1987

    1987

  33. [41]

    Representation learning: A review and new perspectives

    Yoshua Bengio, Aaron Courville, and Pascal Vincent. Representation learning: A review and new perspectives. IEEE transactions on pattern analysis and machine intelligence, 35 0 (8): 0 1798--1828, 2013

    2013

  34. [42]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 770--778, Las Vegas, NV, USA, 2016

    2016

  35. [43]

    How shift equivariance impacts metric learning for instance segmentation

    Josef Lorenz Rumberger, Xiaoyan Yu, Peter Hirsch, Melanie Dohmen, Vanessa Emanuela Guarino, Ashkan Mokarian, Lisa Mais, Jan Funke, and Dagmar Kainmueller. How shift equivariance impacts metric learning for instance segmentation. In Proceedings of the IEEE/CVF International Con...

    2021

  36. [44]

    beta-vae: Learning basic visual concepts with a constrained variational framework

    Irina Higgins, Loic Matthey, Arka Pal, Christopher P Burgess, Xavier Glorot, Matthew M Botvinick, Shakir Mohamed, and Alexander Lerchner. beta-vae: Learning basic visual concepts with a constrained variational framework. ICLR (Poster), 3, 2017

    2017

  37. [45]

    The mnist database of handwritten digit images for machine learning research

    Li Deng. The mnist database of handwritten digit images for machine learning research. IEEE Signal Processing Magazine, 29 0 (6): 0 141--142, 2012

    2012

  38. [46]

    B enchmarking image database for shape recognition techniques

    MPEG -7 C ore E xperiment CE - S hape-1. B enchmarking image database for shape recognition techniques. https://www.ehu.eus/ccwintco/index.php/MPEG-7_Core_Experiment_CE-Shape-1_Test_Set._Benchmarking_image_database_for_shape_recognition_techniques, 2009. Accessed: 2025-01-30

    2009

  39. [47]

    Annotated high-throughput microscopy image sets for validation

    Vebjorn Ljosa, Katherine L Sokolnicki, and Anne E Carpenter. Annotated high-throughput microscopy image sets for validation. Nature Methods, 9 0 (7): 0 637--637, July 2012. ISSN 1548-7091, 1548-7105. doi:10.1038/nmeth.2083. URL http://www.nature.com/articles/nmeth.2083

    2012 doi

  40. [48]

    Phillip, Kyu-Sang Han, Wei-Chiang Chen, Denis Wirtz, and Pei-Hsun Wu

    Jude M. Phillip, Kyu-Sang Han, Wei-Chiang Chen, Denis Wirtz, and Pei-Hsun Wu. A robust unsupervised machine-learning method to quantify the morphological heterogeneity of cells and nuclei. Nature Protocols, 16 0 (2): 0 754--774, 2 2021. ISSN 1750-2799. doi:10.1038/s41596-020-00432-x

    2021 doi

  41. [49]

    Logistic Regression, pages 243--250

    Ekaba Bisong. Logistic Regression, pages 243--250. Apress, Berkeley, CA, 2019. ISBN 978-1-4842-4470-8. doi:10.1007/978-1-4842-4470-8_20. URL https://doi.org/10.1007/978-1-4842-4470-8_20

    2019 doi

  42. [50]

    Chai, Wee Sun Lee, and Hai Leong Chieu

    Nan Ye, Kian Ming A. Chai, Wee Sun Lee, and Hai Leong Chieu. Optimizing f-measures: a tale of two approaches. In Proceedings of the 29th International Conference on International Conference on Machine Learning, ICML'12, page 1555–1562, Edinburgh, Scotland, 2012

    2012

  43. [51]

    Shape distributions

    Robert Osada, Thomas Funkhouser, Bernard Chazelle, and David Dobkin. Shape distributions. ACM Transactions on Graphics (TOG), 21 0 (4): 0 807--832, 2002

    2002

  44. [52]

    Pytorch: An imperative style, high-performance deep learning library

    Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, Alban Desmaison, Andreas Kopf, Edward Yang, Zachary DeVito, Martin Raison, Alykhan Tejani, Sasank Chilamkurthy, Benoit Steiner, Lu ...

    2019

  45. [53]

    Kingma and Jimmy Ba

    Diederik P. Kingma and Jimmy Ba. Adam: A method for stochastic optimization. In Proceedings of the 3rd International Conference on Learning Representations (ICLR), San Diego, CA, USA, 2015. URL https://arxiv.org/abs/1412.6980

    2015 arXiv

  46. [54]

    Mart\' i n Abadi, Ashish Agarwal, Paul Barham, Eugene Brevdo, Zhifeng Chen, Craig Citro, Greg S. Corrado, Andy Davis, Jeffrey Dean, Matthieu Devin, Sanjay Ghemawat, Ian Goodfellow, Andrew Harp, Geoffrey Irving, Michael Isard, Yangqing Jia, Rafal Jozefowicz, Lukasz Kaiser, Manj...

    2015

  47. [55]

    Model assisted survey sampling

    Carl-Erik S \"a rndal, Bengt Swensson, and Jan Wretman. Model assisted survey sampling. Springer Science & Business Media, 2003

    2003

  48. [56]

    Christopher M. Bishop. Pattern Recognition and Machine Learning. Springer, New York, NY, 2006. ISBN 9780387310732. URL https://www.springer.com/gp/book/9780387310732

    2006

  49. [57]

    Visualizing data using t-SNE

    Laurens van der Maaten and Geoffrey Hinton. Visualizing data using t-SNE . Journal of Machine Learning Research, 9: 0 2579--2605, 2008

    2008

  50. [58]

    An image analysis toolbox for high-throughput c

    Carolina W \"a hlby, Lee Kamentsky, Zihan H Liu, Tammy Riklin-Raviv, Annie L Conery, Eyleen J O'rourke, Katherine L Sokolnicki, Orane Visvikis, Vebjorn Ljosa, Javier E Irazoqui, et al. An image analysis toolbox for high-throughput c. elegans assays. Nature methods, 9 0 (7): 0 ...

    2012

  51. [59]

    Smrt analysis of mtoc and nuclear positioning reveals the role of eb1 and lic1 in single-cell polarization

    Christopher M Hale, Wei-Chiang Chen, Shyam B Khatau, Brian R Daniels, Jerry SH Lee, and Denis Wirtz. Smrt analysis of mtoc and nuclear positioning reveals the role of eb1 and lic1 in single-cell polarization. Journal of cell science, 124 0 (24): 0 4267--4285, 2011

    2011

  52. [60]

    Open-source deep-learning software for bioimage segmentation

    Alice M Lucas, Pearl V Ryder, Bin Li, Beth A Cimini, Kevin W Eliceiri, and Anne E Carpenter. Open-source deep-learning software for bioimage segmentation. Molecular Biology of the Cell, 32 0 (9): 0 823--829, 2021

    2021

  53. [61]

    Cellpose: a generalist algorithm for cellular segmentation

    Carsen Stringer, Tim Wang, Michalis Michaelos, and Marius Pachitariu. Cellpose: a generalist algorithm for cellular segmentation. Nature methods, 18 0 (1): 0 100--106, 2021

    2021

  54. [62]

    Oliphant, Matt Haberland, Tyler Reddy, David Cournapeau, Evgeni Burovski, Pearu Peterson, Warren Weckesser, Jonathan Bright, St \'e fan J

    Pauli Virtanen, Ralf Gommers, Travis E. Oliphant, Matt Haberland, Tyler Reddy, David Cournapeau, Evgeni Burovski, Pearu Peterson, Warren Weckesser, Jonathan Bright, St \'e fan J. van der Walt , Matthew Brett, Joshua Wilson, K. Jarrod Millman, Nikolay Mayorov, Andrew R. J. Nels...

    2020

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.