Recognizing Image Objects by Relational Analysis Using Heterogeneous Superpixels and Deep Convolutional Features

Alex Yang; Charlie T. Veal; Derek T. Anderson; Grant J. Scott

REVIEW 4 major objections 5 minor 20 references

Recognizing Image Objects by Relational Analysis Using Heterogeneous Superpixels and Deep Convolutional Features

T0 review · 4 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read A Superpixel Capsule Network pools convolutional features into superpixel regions, routes them as capsules to class capsules, and obtains about 89% validation accuracy on Linnaeus 5 with 88% fewer trainable parameters than VGG-16, while…

desk verdict Genuinely new combination of SLIC superpixels and capsule routing with a defensible feasibility result, but the part-whole interpretability claim is unsupported and the evaluation details are sloppy. read the letter →

arxiv 1908.00669 v1 pith:L3OYST2C submitted 2019-08-02 cs.CV

Alex Yang , Charlie T. Veal , Derek T. Anderson , Grant J. Scott This is my paper · ORCID

classification cs.CV

keywords deepconvolutionalneuralnetworkheterogeneoussuperpixelimageclassificationrelationalanalysiscapsulepoolingvisualexplainability

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

Superpixels are usually treated as a way to compress images by grouping similar pixels, and they are rarely integrated into deep networks so that relationships among object parts are preserved. This paper argues that superpixels can instead supply the structural scaffold of a classifier: a convolutional feature map is pooled into superpixel regions, and each region is treated as a capsule, a vector-valued unit that can be routed toward class-level capsules. The proposed Superpixel Capsule Network learns to classify images by combining these region vectors using dynamic routing, needing only image-level class labels. On the Linnaeus 5 dataset it reaches about 89% validation accuracy while using 88% fewer trainable parameters than the VGG-16 baseline, and it produces per-superpixel contribution maps that highlight the object being classified. If the approach is right, image classification can be made cheaper and more interpretable by building part-whole structure directly into the network.

What carries the argument

The central machinery is a two-stage reorganization. First, heterogeneous superpixels — SLIC regions allowed to have higher internal entropy than conventional oversegmentations — convert a spatial feature map into a structured set of region vectors, preserving the association between features and image locations. Second, dynamic routing treats those region vectors as low-level capsules and iteratively maps them to class-level capsules via $v_j = \varphi(\sum_i c_{ij} W_{ij} u_i)$, where $c_{ij}$ are routing coefficients; the scalar projection $z_{ij} = (W_{ij}u_i)^{\mathsf{T}} v_j / \|v_j\|$ then measures each superpixel's contribution to a class. This combination is what allows the network to learn part-whole relations from classification labels alone and to render those relations as a segmentation-like map.

What would settle it

Replace the SLIC superpixels with random regions of identical size and shape, keeping everything else fixed. If accuracy and the clarity of the contribution maps stay roughly unchanged, the superpixel-feature correspondence is not doing the work claimed; if they drop sharply, the correspondence is confirmed as the mechanism.

Watch

Extended reading notes

Core claim

The central discovery the paper argues for is that superpixel features can be learned with capsules, and that organizing convolutional features by heterogeneous superpixels before relational analysis is beneficial for both recognition and interpretability. A pretrained VGG-16 backbone produces an $8 \times 8 \times 512$ feature map; SLIC superpixels computed on the original image define irregular regions, and each superpixel aggregates the convolutional feature values in its corresponding feature-map tiles, producing one vector per superpixel. These vectors are treated as low-level capsules and routed through learned linear transformations to class capsules whose vector magnitudes indicate class presence. On Linnaeus 5, configurations with roughly 25 to 100 superpixels reach about 88--89% validation accuracy with 17 million trainable parameters, against the VGG-16 baseline's 96% validation accuracy with 134 million parameters. The scalar projection of each superpixel vector onto a class vector can be painted back onto the segmentation, yielding entity contribution maps that align with object parts without any segmentation labels.

Load-bearing premise

The argument rests on the assumption that superpixel regions computed on the full-resolution image align correctly with the coarse $8 \times 8$ convolutional feature map, so each pooled vector represents a genuine spatial part of the object; the paper asserts this correspondence is accurate with its Eq. (5) but runs no experiment that checks the alignment.

Editorial extensions

If this is right

A classifier can be assembled by pooling convolutional features into superpixel regions and routing them with capsules; it needs only class labels, yet it produces region-level explanations of its decision.
Superpixel count acts as a receptive-field hyperparameter: roughly 25--100 superpixels per image keep validation accuracy near 88--89% on Linnaeus 5, while very large superpixels slow convergence and very small ones dilute the contribution maps.
Using 88% fewer trainable parameters than VGG-16, the network reaches about 89% validation accuracy, so strongly reducing parameter count does not force the abandonment of structured part-whole reasoning.
The per-superpixel scalar projections provide object-highlighting maps, meaning segmentation-like interpretability is available without pixel-level or superpixel-level labels.
Widening the class vector from 16 to 64 dimensions lowers accuracy across the board in the reported experiments, indicating that simply increasing capsule capacity is not an effective route to better performance.

Reading between the lines

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

If the superpixel-to-feature-map correspondence is sound, the same relational-analysis layer could be attached to lighter convolutional backbones than VGG-16, trading a little accuracy for a much smaller parameter budget with built-in interpretability.
The contribution maps could serve as a weak-supervision signal: thresholds applied to the per-superpixel projections might generate pseudo-segmentation masks with which to train other models, something the paper does not test.
The consistent drop with 64-dimensional class vectors suggests capsule routing may need more data or regularization when the routing space expands; a direct test would be to train longer on a larger dataset and see whether larger vectors recover.
The reported empirical relation that mean superpixel entropy scales roughly as the inverse of superpixel count could be used to choose superpixel granularity from image resolution alone, before training multiple models.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Two things to know about arXiv:1908.00669. First, the architecture is genuinely new: nobody else pools VGG-16 conv features into SLIC superpixels and then runs capsule routing over them, and the paper earns credit for the attempt. Second, the central feasibility claim holds up — you can get roughly 89% validation accuracy on Linnaeus 5 with far fewer parameters than VGG-16 — but the stronger claim about learning part-whole relations is not supported by the experiments as reported.

What it does well: the 'heterogeneous superpixel' framing is a useful counterpoint to the usual homogeneity assumption, and Table III shows consistent accuracy across many superpixel counts, which suggests the result is not a fluke. Replacing the dense layers with a capsule layer cuts trainable parameters by 88%, and that is a real contribution. The contribution maps from Eq. (10) are a nice interpretability device, and the figures show plausible object highlighting.

The soft spots are real but not all equally soft. The dataset is described inconsistently: 5 categories in one place, 4 classes plus adversarial examples elsewhere, and 'each of the 5 categories contains 1200 training images' does not square with the 4-class/6400-image description used for the VGG-16 comparison. There are no error bars despite claiming 10-fold cross-validation. The 89% number is below plain VGG-16's 96% on the same data, so the capsule layer is not an accuracy win — fine, the claim is about compactness and interpretability, but the paper should say that more clearly.

The more serious flaw is the receptive field problem. Eqs. (5)-(6) map each 8x8 conv5 feature vector to a 32x32 tile at the original resolution, but VGG-16 conv5 features have an effective receptive field of roughly 196x196 pixels. So each feature vector encodes not just its tile but a large surrounding context. The paper asserts the association is 'accurate' but gives no analysis or experiment to justify that the superpixel-pooled features actually correspond to localized object parts. That means the interpretation of Figs. 7-8 as learned part-whole segmentations is speculative. The classification result does not depend on that interpretation, so the feasibility claim survives, but the interpretability claim is overstated.

Who this is for: readers working on capsule networks, superpixel pooling, or weakly supervised interpretability. It deserves a serious referee — the idea is worth the time — but the authors should be asked to fix the dataset description, add error bars, and either provide an experiment validating the feature-to-superpixel correspondence or soften the part-whole language.

Referee Report

4 major / 5 minor

Summary. The paper proposes a Superpixel Capsule Network for image classification. A VGG-16 network (pretrained on ImageNet) produces 8×8×512 convolutional feature maps, which are reorganized by pooling into a variable number of SLIC superpixels (Eqs. (4)–(7)). The resulting S×k superpixel feature vectors are treated as low-level capsules and routed into class capsules via dynamic routing (Eqs. (8)–(9)). The magnitude of each class vector gives a classification score, and a scalar projection (Eq. (10)) is used to produce superpixel-wise contribution maps for visual explanation. The authors report about 89% validation accuracy on the Linnaeus 5 dataset with roughly 88% fewer trainable parameters than VGG-16, and they argue that the resulting part-whole structure provides interpretability and a basis for localization without pixel-level labels.

Significance. If the claims are upheld, the architecture offers a parameter-efficient alternative to standard dense classification heads and provides an intuitive form of explainability through superpixel contribution maps. The paper combines standard components (VGG-16 features, SLIC, capsule routing) in a novel pipeline and evaluates it on a public dataset. The consistent accuracy across many superpixel counts (Table III) is a positive indicator. However, the significance of the interpretability claim depends critically on whether the superpixel-feature association preserves spatial part-whole relations, which the paper does not validate. The evaluation also lacks statistical rigor (no error bars, inconsistent dataset description), and the benefit of capsule routing over simpler superpixel-based classifiers is not demonstrated.

major comments (4)

[§II-B, Eqs. (5)–(6)] The association of each conv5 feature vector with a 32×32 image tile is not sufficient to make the feature vector a localized descriptor of that tile. In VGG-16, the effective receptive field of conv5_3 features is roughly 196×196 pixels at the original resolution, so each feature vector encodes content far outside its assigned tile. Consequently, superpixel features aggregated via Eq. (7) are mixtures of overlapping large-context descriptors, not localized representations of the superpixel regions. The claim in §II-B that the association 'is accurate with Eq. 5' is only a statement about tiling; it does not address the receptive-field mismatch. This undermines the central interpretability claims in §V-B and Figs. 7–8, where the network's internal part-whole representation is inferred from superpixel contribution maps. The authors should provide an experiment or analysis that validates the spatial locality of the features used for superpixel pooling, or soften the part-whole interpretation accordingly.
[§V, dataset description] The dataset description is internally inconsistent. Section V states that Linnaeus 5 'has 4 classes: berry, bird, dog and flower and other unclassified images as adversarial examples,' but two sentences later says 'Each of the 5 categories contains 1200 training images and 400 test images.' Section V-C then refers to 'the 4-class dataset consisting of 6400 images' with 25% held out as validation. These statements cannot all be true, and they prevent a reader from determining what was actually trained and evaluated. The paper must clarify the exact number of classes, the number of images per class, and the train/test or cross-validation split actually used.
[§V, Table III] The paper claims a 10-fold cross-validation methodology, but Table III reports only a single training and validation accuracy for each configuration, with no standard deviations or confidence intervals. Given the small differences among S=25, 36, 50, and 100 (all around 88–89%), it is impossible to assess whether the observed patterns are statistically meaningful. The authors should report the mean and standard deviation across folds (or across repeated runs) for the central configurations, and state whether the numbers in Table III correspond to a single fold, the average, or the best run. Without this, the empirical support for the main feasibility claim is incomplete.
[§III, §V] The paper does not include a baseline that uses the same superpixel feature extraction and class labels but replaces capsule routing with a simpler classifier (e.g., a fully connected layer or a global average of superpixel features followed by logistic regression). Such a baseline is needed to support the claim that relational analysis via capsules is 'potentially beneficial' over naive aggregation. As presented, the results show only that superpixel features plus capsules achieve about 89% accuracy; they do not isolate the contribution of the capsule routing mechanism. Adding this baseline would make the central claim about the value of relational analysis testable.

minor comments (5)

[General] There are numerous typographical errors, including 'Supixelpixel' in the captions of Figs. 7 and 8, 'segmemtations' after Eq. (10), and 'backberry' in §V-B. The paper would benefit from a careful proofreading pass.
[Eq. (2)] The entropy operator H(·) is used in Eqs. (2)–(3) and in the text but is never formally defined. The text mentions a 256-bin hue histogram with respect to each superpixel region; this definition should be made explicit in a formal equation or at the point of first use.
[Table II] The layout of Table II is confusing: the column headers appear misaligned, and the 'M' values implied by Eq. (3) are not listed. It would be clearer to present the superpixel-weighted entropy directly and state the fitted value of M for each row.
[§IV, training setup] The learning rate, batch size, and optimizer settings are only partially specified (stochastic gradient descent with learning rate 2×10^{-5} is stated, but no batch size or momentum is given). These details are needed for reproducibility.
[References] Reference [1] is cited as the source of both ImageNet and the pretrained VGG-16 weights, but it is a URL to the ILSVRC challenge page, not a citable paper. The authors should cite the ImageNet publication and the VGG-16 paper more precisely.

Circularity Check

1 steps flagged · score 1.0 of 10

Minor self-definitional step in the resolution-association proof; the classification claim itself is empirical and not circular.

self definitional [Section II-B, Eqs. (5)-(6)]
"The value of any pixel in an image ( x :V → R) up-scaled ... by an integer factor (to x′ :V → R = x◦τ) can be found directly from x. ... ∀p∈Tij : (x′◦T )(i,j ) = x′(Tij) = x′(p) = (x◦τ )(p) = x(τ (p)) = x(i,j ) . (5) in short, x =x′◦T . (6) ... Such direct association is pointed out in [5] as well, but it is demonstrated that it is accurate with Eq. 5."

Eq. (5) is true by construction: x′ is defined as x∘τ, so the tile-constant equality x = x′∘T restates the definition rather than proving accuracy. The paper says this demonstrates that the direct association across feature-map resolutions 'is accurate', but no experiment or analytic argument shows that VGG-16 conv5 features are localized to their assigned tiles; their receptive fields are much larger than one tile. The association is therefore an unvalidated modeling assumption presented as a derivation. This is a self-definitional step, not a circular derivation of the classification numbers, which come from direct training and validation.

full rationale

The central claim is empirical: a specific Superpixel Capsule Network with VGG-16 features was trained on Linnaeus 5 and evaluated against VGG-16 (Table III, Fig. 9). The pooling equations (4) and (7), capsule routing (8), and contribution projection (10) are standard constructions, not fitted parameters renamed as predictions. The entropy relation M≈S^{-1} is an empirical observation about superpixel segmentation and is not used to predict classification accuracy. The paper contains no load-bearing self-citation: its references to SLIC, VGG-16, and capsule routing are external works, and none of the defining theorems or algorithms are imported from the present authors' prior publications. The only reduction-by-construction is the resolution-association proof in Eqs. (5)-(6), which asserts accuracy by definition rather than by measurement; this is a minor logical gap in the interpretability argument, not a circularity in the classification derivation. The receptive-field concern raised by the skeptic is a correctness risk for the part-whole interpretation, not a circularity. Score 1.

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

No new physical or mathematical entities are introduced; the 'heterogeneous superpixel' is a configuration choice, not an invention. The central claim rests primarily on the structural assumption that superpixels align with object parts and on the empirical effectiveness of the architecture.

free parameters (5)

Number of superpixels S = 10, 16, 25, 36, 50, 100, 200 (swept)
Chosen by hand; the paper identifies it as the hyperparameter with the greatest impact on accuracy and convergence (Sect. IV, Table III). The central feasibility claim is demonstrated across this range, so the claim does not hinge on a single value, but the optimal value is data-dependent.
Capsule class vector dimension Q = 16 and 64
Chosen by hand; Table III shows Q=16 yields consistently higher accuracy than Q=64. The paper uses Q=16 for the headline comparison.
SLIC compactness m = 0.1
Set for all experiments (Sect. IV-A); controls superpixel shape irregularity.
SLIC smoothing factor sigma = 0
Set for all experiments (Sect. IV-A); controls smoothing in SLIC.
Entropy scaling relation M approximately S^{-1} = empirically fit to Table II
Empirical relation used to argue that superpixel pooling reduces local entropy more efficiently than convolution; it is fit to the paper's own entropy measurements and is not a derived law, but it is not central to the classification claim.

assumptions (4)

domain assumption Superpixels at full resolution can be mapped onto lower-resolution conv feature maps via Eqs. (5)-(6) without loss of spatial association.
The conv feature map is 8x8 while the segmentation is 256x256; the paper claims direct association is accurate but provides no validation.
domain assumption Pretrained VGG-16 features are an adequate representation for the subsequent superpixel pooling and capsule layers.
The architecture relies on transfer learning from ImageNet; the paper does not test alternative feature extractors.
ad hoc to paper Hue histogram entropy over superpixel regions is a meaningful measure of information content for this task.
Used to motivate 'heterogeneous superpixels' (Fig. 1, Tab. II); no justification for using hue only rather than the full feature distribution.
domain assumption Standard capsule dynamic routing generalizes to inputs that are superpixel feature vectors of varying spatial structure.
The paper treats each superpixel as a lower-level capsule without modifying the routing algorithm; this may be reasonable but is unexamined.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Recognizing Image Objects by Relational Analysis Using Heterogeneous Superpixels and Deep Convolutional Features." pith.science (2026). https://pith.science/paper/L3OYST2C

@misc{pith2026190800669,
  author       = {Pith},
  title        = {Pith review of: Recognizing Image Objects by Relational Analysis Using Heterogeneous Superpixels and Deep Convolutional Features},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/L3OYST2C}},
  note         = {Machine review of arXiv:1908.00669}
}

read the original abstract

Superpixel-based methodologies have become increasingly popular in computer vision, especially when the computation is too expensive in time or memory to perform with a large number of pixels or features. However, rarely is superpixel segmentation examined within the context of deep convolutional neural network architectures. This paper presents a novel neural architecture that exploits the superpixel feature space. The visual feature space is organized using superpixels to provide the neural network with a substructure of the images. As the superpixels associate the visual feature space with parts of the objects in an image, the visual feature space is transformed into a structured vector representation per superpixel. It is shown that it is feasible to learn superpixel features using capsules and it is potentially beneficial to perform image analysis in such a structured manner. This novel deep learning architecture is examined in the context of an image classification task, highlighting explicit interpretability (explainability) of the network's decision making. The results are compared against a baseline deep neural model, as well as among superpixel capsule networks with a variety of hyperparameter settings.

Figures

Figures reproduced from arXiv: 1908.00669 by the authors.

**Figure 2.** Feature space reorganization using superpixel: On the top right corner [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗

**Figure 3.** Visualization of relational analysis. This figure is broken up into [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Figures from the paper (6 more)

**Figure 4.** Figure 4: An example of entity segmentation using superpixel feature contribu [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]

**Figure 5.** Figure 5: Proposed Network architecture: Masking is used to selectively train [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]

**Figure 6.** Figure 6: Comparison of training accuracy and loss between different superpixel [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]

**Figure 7.** Figure 7: Supixelpixel-wise contribution to the entity feature vector visualized: on the left most column are input images; the rest of the columns (from left to [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]

**Figure 8.** Figure 8: Supixelpixel-wise contribution to the class vector visualized with the sample sample across many different superpixel sizes. The approximated [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]

**Figure 9.** Figure 9: Comparison of training accuracy and loss between Superpixel Capsule [PITH_FULL_IMAGE:figures/full_fig_p010_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

20 extracted references · 16 canonical work pages

[1]

Ilsvrc-2012,

J Deng, A Berg, S Satheesh, H Su, A Khosla, and L Fei-Fei. Ilsvrc-2012,

work page 2012
[2]

Microsoft coco: Common objects in context

Tsung-Yi Lin, Michael Maire, Serge Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Doll´ar, and C Lawrence Zitnick. Microsoft coco: Common objects in context. In European conference on computer vision, pages 740–755. Springer, 2014

2014
[3]

Fully convolutional networks for semantic segmentation

Jonathan Long, Evan Shelhamer, and Trevor Darrell. Fully convolutional networks for semantic segmentation. In Proceedings of the IEEE conference on computer vision and pattern recognition , pages 3431– 3440, 2015

2015
[4]

Understanding convolution for semantic segmentation

Panqu Wang, Pengfei Chen, Ye Yuan, Ding Liu, Zehua Huang, Xiaodi Hou, and Garrison Cottrell. Understanding convolution for semantic segmentation. In 2018 IEEE Winter Conference on Applications of Computer Vision (WACV), pages 1451–1460. IEEE, 2018

work page 2018
[5]

F. Liu, C. Shen, G. Lin, and I. Reid. Learning depth from single monocular images using deep convolutional neural ﬁelds. IEEE Trans. Pattern Anal. Mach. Intell. , 38(10), oct 2016

work page 2016
[6]

Deep learning advances in computer vision with 3d data: A survey

Anastasia Ioannidou, Elisavet Chatzilari, Spiros Nikolopoulos, and Ioan- nis Kompatsiaris. Deep learning advances in computer vision with 3d data: A survey. ACM Computing Surveys (CSUR) , 50(2):20, 2017

work page 2017
[7]

Deformable convnets v2: More deformable, better results

Xizhou Zhu, Han Hu, Stephen Lin, and Jifeng Dai. Deformable convnets v2: More deformable, better results. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages 9308– 9316, 2019

work page 2019
[8]

Ef- ﬁcient semantic image segmentation with superpixel pooling

Mathijs Schuurmans, Maxim Berman, and Matthew B Blaschko. Ef- ﬁcient semantic image segmentation with superpixel pooling. arXiv preprint arXiv:1806.02705, 2018

arXiv 2018

Show all 20 references

[9]

Achanta, A

R. Achanta, A. Shaji, K. Smith, A. Lucchi, P. Fua, and S. Susstrunk. Slic superpixels compared to state-of-the-art superpixel methods. IEEE Trans. Pattern Anal. Mach. Intell. , 34(11):22742282, nov 2012

2012
[10]

Fast cloud image segmentation with superpixel analysis based convolutional networks

Lifang Wu, Jiaoyu He, Meng Jian, Jianan Zhang, and Yunzhen Zou. Fast cloud image segmentation with superpixel analysis based convolutional networks. In Systems, Signals and Image Processing (IWSSIP), 2017 International Conference on , pages 1–5. IEEE, 2017

2017
[11]

Feedforward semantic segmentation with zoom-out features

Mohammadreza Mostajabi, Payman Yadollahpour, and Gregory Shakhnarovich. Feedforward semantic segmentation with zoom-out features. In Proceedings of the IEEE conference on computer vision and pattern recognition , pages 3376–3385, 2015

2015
[12]

Dynamic routing between capsules

Sara Sabour, Nicholas Frosst, and Geoffrey E Hinton. Dynamic routing between capsules. In Advances in neural information processing systems, pages 3856–3866, 2017

2017
[13]

U-net: Convo- lutional networks for biomedical image segmentation

Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-net: Convo- lutional networks for biomedical image segmentation. In International Conference on Medical image computing and computer-assisted inter- vention, pages 234–241. Springer, 2015

2015
[14]

Capsdemm: Capsule network for detection of munros microabscess in skin biopsy images

Anabik Pal, Akshay Chaturvedi, Utpal Garain, Aditi Chandra, Raghu- nath Chatterjee, and Swapan Senapati. Capsdemm: Capsule network for detection of munros microabscess in skin biopsy images. In International Conference on Medical Image Computing and Computer- Assisted Interven...

2018
[15]

Capsules for object segmentation

Rodney LaLonde and Ulas Bagci. Capsules for object segmentation. arXiv preprint arXiv:1804.04241 , 2018

2018 arXiv
[16]

gslicr: Slic superpixels at over 250hz

Carl Yuheng Ren, Victor Adrian Prisacariu, and Ian D Reid. gslicr: Slic superpixels at over 250hz. arXiv preprint arXiv:1509.04232 , 2015

2015 arXiv
[17]

Matrix capsules with em routing

Geoffrey E Hinton, Sara Sabour, and Nicholas Frosst. Matrix capsules with em routing. 6th International Conference on Learning Represen- tations, ICLR, 2018

2018
[18]

Simonyan and A

K. Simonyan and A. Zisserman. Very deep convolutional networks for large-scale image recognition. In Int. Conf. Learn. Representations , 2015

2015
[19]

Linnaeus 5 dataset for machine learning

Giorgi Chaladze and Levan kalatozishvili. Linnaeus 5 dataset for machine learning. 2017

2017
[2012]

image-net

URL http://www. image-net. org/challenges/LSVRC , 2012

2012

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

Ilsvrc-2012,

J Deng, A Berg, S Satheesh, H Su, A Khosla, and L Fei-Fei. Ilsvrc-2012,

work page 2012

[2] [2]

Microsoft coco: Common objects in context

Tsung-Yi Lin, Michael Maire, Serge Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Doll´ar, and C Lawrence Zitnick. Microsoft coco: Common objects in context. In European conference on computer vision, pages 740–755. Springer, 2014

2014

[3] [3]

Fully convolutional networks for semantic segmentation

Jonathan Long, Evan Shelhamer, and Trevor Darrell. Fully convolutional networks for semantic segmentation. In Proceedings of the IEEE conference on computer vision and pattern recognition , pages 3431– 3440, 2015

2015

[4] [4]

Understanding convolution for semantic segmentation

Panqu Wang, Pengfei Chen, Ye Yuan, Ding Liu, Zehua Huang, Xiaodi Hou, and Garrison Cottrell. Understanding convolution for semantic segmentation. In 2018 IEEE Winter Conference on Applications of Computer Vision (WACV), pages 1451–1460. IEEE, 2018

work page 2018

[5] [5]

F. Liu, C. Shen, G. Lin, and I. Reid. Learning depth from single monocular images using deep convolutional neural ﬁelds. IEEE Trans. Pattern Anal. Mach. Intell. , 38(10), oct 2016

work page 2016

[6] [6]

Deep learning advances in computer vision with 3d data: A survey

Anastasia Ioannidou, Elisavet Chatzilari, Spiros Nikolopoulos, and Ioan- nis Kompatsiaris. Deep learning advances in computer vision with 3d data: A survey. ACM Computing Surveys (CSUR) , 50(2):20, 2017

work page 2017

[7] [7]

Deformable convnets v2: More deformable, better results

Xizhou Zhu, Han Hu, Stephen Lin, and Jifeng Dai. Deformable convnets v2: More deformable, better results. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages 9308– 9316, 2019

work page 2019

[8] [8]

Ef- ﬁcient semantic image segmentation with superpixel pooling

Mathijs Schuurmans, Maxim Berman, and Matthew B Blaschko. Ef- ﬁcient semantic image segmentation with superpixel pooling. arXiv preprint arXiv:1806.02705, 2018

arXiv 2018