Linking Art through Human Poses

Ond\v{r}ej Chum; Tomas Jenicek

arxiv: 1907.03537 · v1 · pith:HMVJ5QM6new · submitted 2019-07-08 · 💻 cs.CV

Linking Art through Human Poses

Tomas Jenicek , Ond\v{r}ej Chum This is my paper

Pith reviewed 2026-05-25 01:07 UTC · model grok-4.3

classification 💻 cs.CV

keywords art composition transferhuman pose matchingimage retrievalart history analysiscultural heritagepose similarityspatial verification

0 comments

The pith

Matching human poses in artworks discovers composition transfers better than standard image retrieval.

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

The paper shows that linking artworks through the similarity of human poses they contain finds cases of composition transfer that usual visual matching misses. Artists have long reused distinctive poses as a way to transfer ideas across works, and this holds from the middle ages onward. The method first matches poses quickly then verifies their spatial arrangement. On a dataset of annotated art examples, this explicit pose approach beats content-based retrieval. The result matters for scanning growing digitized museum collections to trace artistic influences automatically.

Core claim

The paper claims that explicit human pose matching is superior to standard content-based image retrieval methods on a manually annotated art composition transfer dataset. The approach consists of two steps: fast pose matching and robust spatial verification. Human figures are the subject of a large fraction of visual art and their distinctive poses were often a source of inspiration among artists, so pose similarity serves as a direct signal for composition transfer that visual similarity alone does not capture.

What carries the argument

Two-step pipeline of fast pose matching followed by robust spatial verification that links artworks via similarity of depicted human figures.

If this is right

Composition transfers hidden from visual similarity searches become detectable through pose links.
Large art collections can be scanned for artistic influence using pose as the connecting signal.
The annotated dataset provides a benchmark for evaluating pose-based retrieval in art.
Spatial verification after initial pose matching reduces false matches in the results.

Where Pith is reading between the lines

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

The same pose signal could be combined with other cues such as color palettes or object types to strengthen links.
Extending the method to non-human subjects like animals or architecture might reveal other transfer patterns.
Networks built from pose matches could quantify how influence spreads across time periods or regions.
Applying the approach to contemporary art or non-European traditions would test whether the pose focus holds outside the paper's scope.

Load-bearing premise

Human figures and their poses form a major source of composition transfer across visual art.

What would settle it

Run the pose-matching method and standard retrieval on the full annotated dataset and find that the pose method does not return more of the expert-labeled composition transfers.

Figures

Figures reproduced from arXiv: 1907.03537 by Ond\v{r}ej Chum, Tomas Jenicek.

**Figure 2.** Figure 2: An example of discovered relations linking artworks in different [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗

**Figure 3.** Figure 3: A visualization of 2D poses output by the OpenPose detec [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗

**Figure 4.** Figure 4: Examples of discovered links between artworks in the WGA database that are impossible to retrieve using standard image retrieval methods [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗

**Figure 5.** Figure 5: An example of false positives retrieved by our method: incorrect [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗

**Figure 6.** Figure 6: Failure cases of the OpenPose detector. In the left image, only the [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗

read the original abstract

We address the discovery of composition transfer in artworks based on their visual content. Automated analysis of large art collections, which are growing as a result of art digitization among museums and galleries, is an important tool for art history and assists cultural heritage preservation. Modern image retrieval systems offer good performance on visually similar artworks, but fail in the cases of more abstract composition transfer. The proposed approach links artworks through a pose similarity of human figures depicted in images. Human figures are the subject of a large fraction of visual art from middle ages to modernity and their distinctive poses were often a source of inspiration among artists. The method consists of two steps -- fast pose matching and robust spatial verification. We experimentally show that explicit human pose matching is superior to standard content-based image retrieval methods on a manually annotated art composition transfer dataset.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note private letter to a colleague

Pose matching beats generic CBIR for art composition transfer on their annotated set, but the method's reliance on off-the-shelf pose detectors trained on photos is untested in the target domain.

read the letter

The paper shows that running human pose estimation, then matching and verifying those poses, finds composition transfers in art better than standard content-based retrieval on a manually labeled test set. That is the concrete result they report. It is a targeted application rather than a new algorithm or theory. The motivation is sound: many artworks from the middle ages onward center on human figures, and artists have long copied or varied poses from earlier works. Framing retrieval around pose similarity therefore aligns with how art historians actually look for influence. The two-step pipeline (fast pose match followed by spatial verification) is straightforward and avoids overclaiming generality. They also built and annotated a dataset for this specific task, which is useful even if small. That is the part worth crediting. The main gap is the missing check on whether the pose detector itself works on the art images. Standard estimators are trained on photographs; paintings and drawings introduce line style, abstraction, and non-photorealistic proportions that often degrade keypoint accuracy. The abstract gives no separate pose-estimation numbers on the art collection, no failure cases, and no ablation that isolates the contribution of pose quality from the verification stage. Without those, it is hard to know whether the reported gain is robust or partly an artifact of the test set. The circularity risk is low because the final metric is retrieval performance against human annotations, not self-referential. Still, the domain-shift issue is the load-bearing assumption that needs numbers. This work is aimed at the small group doing computer vision for cultural heritage collections. A reader already working on art retrieval or pose transfer would find the dataset and the direct comparison useful. It is not broad enough or methodologically novel enough to interest most vision researchers outside that niche. I would send it to peer review so the authors can supply the pose accuracy diagnostics and any error analysis they have. The central claim is narrow enough that a referee can evaluate it on the evidence they provide rather than on speculation.

Referee Report

2 major / 1 minor

Summary. The paper proposes linking artworks through explicit human pose similarity for discovering composition transfers. It describes a two-step method of fast pose matching followed by robust spatial verification and reports experimental superiority over standard content-based image retrieval on a manually annotated art composition transfer dataset.

Significance. If the results hold, the work offers a targeted retrieval approach for art history applications, exploiting the prevalence of human figures and poses across periods as a basis for identifying influences that generic visual similarity methods miss.

major comments (2)

[Abstract, §3] Abstract and method description: the central claim of superiority for explicit pose matching requires reliable pose detections on artistic images, yet no quantitative pose-estimation accuracy, error analysis, or ablation isolating pose quality from spatial verification is provided on the target non-photorealistic dataset.
[Experimental section] Experimental evaluation: the manually annotated test set is central to the superiority claim, but insufficient detail is given on dataset size, annotation protocol, inter-annotator agreement, or how composition-transfer ground truth was established, preventing assessment of whether results generalize or are dataset-specific.

minor comments (1)

[Abstract] The abstract states the motivation regarding human figures in art but does not quantify the fraction of artworks containing detectable figures or discuss failure cases when figures are absent or heavily abstracted.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback highlighting areas where additional detail would strengthen the paper. We address each major comment below and indicate planned revisions.

read point-by-point responses

Referee: [Abstract, §3] Abstract and method description: the central claim of superiority for explicit pose matching requires reliable pose detections on artistic images, yet no quantitative pose-estimation accuracy, error analysis, or ablation isolating pose quality from spatial verification is provided on the target non-photorealistic dataset.

Authors: We agree that quantitative pose-estimation accuracy on artistic images and an ablation isolating the contribution of pose quality would strengthen the central claim. The current experiments focus on end-to-end retrieval performance. We will add qualitative examples of pose detections on the art dataset and an ablation comparing pose matching with and without spatial verification in the revised manuscript. A full quantitative benchmark would require new ground-truth pose annotations on the target dataset, which we do not currently possess. revision: partial
Referee: [Experimental section] Experimental evaluation: the manually annotated test set is central to the superiority claim, but insufficient detail is given on dataset size, annotation protocol, inter-annotator agreement, or how composition-transfer ground truth was established, preventing assessment of whether results generalize or are dataset-specific.

Authors: We will expand the experimental section with the requested details: exact dataset size (number of images and annotated composition transfers), the annotation protocol, how ground truth was established, and any available inter-annotator agreement statistics. These additions will clarify the evaluation setup and support assessment of generalizability. revision: yes

Circularity Check

0 steps flagged

No circularity; method and evaluation are independent of inputs

full rationale

The paper presents an empirical method consisting of pose matching followed by spatial verification, then reports experimental superiority on a manually annotated dataset. No derivation chain, fitted parameters renamed as predictions, or self-citation load-bearing steps are present in the provided text. The central claim rests on direct comparison to CBIR baselines rather than reducing to its own assumptions or prior self-citations by construction. The assumption that human poses are common in art is stated explicitly but does not create a definitional loop with the matching procedure.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The approach rests on the domain assumption that human poses are a reliable signal for composition transfer in art; no free parameters or invented entities are described in the abstract.

axioms (1)

domain assumption Human figures are the subject of a large fraction of visual art from middle ages to modernity and their distinctive poses were often a source of inspiration among artists.
Invoked in the abstract to justify focusing on pose similarity rather than general visual content.

pith-pipeline@v0.9.0 · 5658 in / 1117 out tokens · 18754 ms · 2026-05-25T01:07:40.090595+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

29 extracted references · 29 canonical work pages · 3 internal anchors

[1]

New techniques for the digitization of art historical photographic archives-the case of the cini foundation in venice,

B. Seguin, L. Costiner, I. di Lenardo, and F. Kaplan, “New techniques for the digitization of art historical photographic archives-the case of the cini foundation in venice,” in Archiv- ing Conference , vol. 2018, no. 1. Society for Imaging Science and Technology, 2018, pp. 1–5

work page 2018
[2]

Mas- sive 3d digitization of museum contents,

G. Guidi, S. G. Barsanti, L. L. Micoli, and M. Russo, “Mas- sive 3d digitization of museum contents,” in Built heritage: Monitoring conservation management . Springer, 2015, pp. 335–346

work page 2015
[3]

Visual link retrieval in a database of paintings,

B. Seguin, C. Striolo, F. Kaplan et al., “Visual link retrieval in a database of paintings,” in European Conference on Computer Vision. Springer, 2016, pp. 753–767

work page 2016
[4]

Tracking transmis- sion of details in paintings

B. Seguin, I. diLenardo, and F. Kaplan, “Tracking transmis- sion of details in paintings.” in DH, 2017

work page 2017
[5]

The replica project: building a visual search engine for art historians,

B. Seguin, “The replica project: building a visual search engine for art historians,” XRDS: Crossroads, The ACM Magazine for Students , vol. 24, no. 3, pp. 24–29, 2018

work page 2018
[6]

Artistic image classiﬁcation: An analysis on the printart database,

G. Carneiro, N. P. da Silva, A. Del Bue, and J. P. Costeira, “Artistic image classiﬁcation: An analysis on the printart database,” in European Conference on Computer Vision . Springer, 2012, pp. 143–157

work page 2012
[7]

Classiﬁcation of artistic styles using binarized features derived from a deep neural network,

Y . Bar, N. Levy, and L. Wolf, “Classiﬁcation of artistic styles using binarized features derived from a deep neural network,” in Workshop at the European Conference on Computer Vision. Springer, 2014, pp. 71–84

work page 2014
[8]

Large-scale Classification of Fine-Art Paintings: Learning The Right Metric on The Right Feature

B. Saleh and A. Elgammal, “Large-scale classiﬁcation of ﬁne- art paintings: Learning the right metric on the right feature,” arXiv preprint arXiv:1505.00855 , 2015

work page internal anchor Pith review Pith/arXiv arXiv 2015
[9]

Towards a hierarchical multitask classiﬁcation framework for cultural heritage,

A. Belhi, A. Bouras, and S. Foufou, “Towards a hierarchical multitask classiﬁcation framework for cultural heritage,” in 2018 IEEE/ACS 15th International Conference on Computer Systems and Applications (AICCSA) . IEEE, 2018, pp. 1–7

work page 2018
[10]

Detecting people in artwork with cnns,

N. Westlake, H. Cai, and P. Hall, “Detecting people in artwork with cnns,” in European Conference on Computer Vision . Springer, 2016, pp. 825–841

work page 2016
[11]

Weakly supervised object detection in artworks,

N. Gonthier, Y . Gousseau, S. Ladjal, and O. Bonfait, “Weakly supervised object detection in artworks,” in European Con- ference on Computer Vision . Springer, 2018, pp. 692–709

work page 2018
[12]

How to read paintings: Semantic art understanding with multi-modal retrieval,

N. Garcia and G. V ogiatzis, “How to read paintings: Semantic art understanding with multi-modal retrieval,” in European Conference on Computer Vision . Springer, 2018, pp. 676– 691

work page 2018
[13]

The rijksmuseum challenge: Museum-centered visual recognition,

T. Mensink and J. Van Gemert, “The rijksmuseum challenge: Museum-centered visual recognition,” in Proceedings of In- ternational Conference on Multimedia Retrieval . ACM, 2014, p. 451

work page 2014
[14]

Recognizing image style,

S. Karayev, M. Trentacoste, H. Han, A. Agarwala, T. Darrell, A. Hertzmann, and H. Winnemoeller, “Recognizing image style,” arXiv preprint arXiv:1311.3715 , 2013

work page arXiv 2013
[15]

The state of the art: Object retrieval in paintings using discriminative regions

E. Crowley and A. Zisserman, “The state of the art: Object retrieval in paintings using discriminative regions.” in BMVC, 2014

work page 2014
[16]

In search of art,

E. J. Crowley and A. Zisserman, “In search of art,” in Workshop at the European Conference on Computer Vision . Springer, 2014, pp. 54–70

work page 2014
[17]

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

J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, “You only look once: Uniﬁed, real-time object detection,” in Pro- ceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 779–788

work page 2016
[18]

The Cross-Depiction Problem: Computer Vision Algorithms for Recognising Objects in Artwork and in Photographs

H. Cai, Q. Wu, T. Corradi, and P. Hall, “The cross- depiction problem: Computer vision algorithms for recognis- ing objects in artwork and in photographs,” arXiv preprint arXiv:1505.00110, 2015

work page internal anchor Pith review Pith/arXiv arXiv 2015
[19]

Data-driven visual similarity for cross-domain image match- ing,

A. Shrivastava, T. Malisiewicz, A. Gupta, and A. A. Efros, “Data-driven visual similarity for cross-domain image match- ing,” ACM Transactions on Graphics (ToG) , vol. 30, no. 6, p. 154, 2011

work page 2011
[20]

Deep shape match- ing,

F. Radenovic, G. Tolias, and O. Chum, “Deep shape match- ing,” in ECCV, 2018

work page 2018
[21]

Pose search: retrieving people using their pose,

V . Ferrari, M. Marin-Jimenez, and A. Zisserman, “Pose search: retrieving people using their pose,” in Computer Vision and Pattern Recognition, 2009. CVPR 2009. IEEE Conference on. IEEE, 2009, pp. 1–8

work page 2009
[22]

2d articulated human pose estimation and retrieval in (almost) unconstrained still images,

M. Eichner, M. Marin-Jimenez, A. Zisserman, and V . Fer- rari, “2d articulated human pose estimation and retrieval in (almost) unconstrained still images,” International journal of computer vision, vol. 99, no. 2, pp. 190–214, 2012

work page 2012
[23]

Visual sentences for pose retrieval over low-resolution cross-media dance collections,

R. Ren and J. Collomosse, “Visual sentences for pose retrieval over low-resolution cross-media dance collections,” IEEE Transactions on Multimedia , vol. 14, no. 6, pp. 1652–1661, 2012

work page 2012
[24]

OpenPose: Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields

Z. Cao, G. Hidalgo, T. Simon, S.-E. Wei, and Y . Sheikh, “OpenPose: realtime multi-person 2D pose estimation using Part Afﬁnity Fields,” in arXiv preprint arXiv:1812.08008 , 2018

work page internal anchor Pith review Pith/arXiv arXiv 2018
[25]

RMPE: Regional multi-person pose estimation,

H.-S. Fang, S. Xie, Y .-W. Tai, and C. Lu, “RMPE: Regional multi-person pose estimation,” in ICCV, 2017

work page 2017
[26]

Mask r- cnn,

K. He, G. Gkioxari, P. Doll ´ar, and R. Girshick, “Mask r- cnn,” in Computer Vision (ICCV), 2017 IEEE International Conference on. IEEE, 2017, pp. 2980–2988

work page 2017
[27]

Random sample consensus: a paradigm for model ﬁtting with applications to image analysis and automated cartography,

M. A. Fischler and R. C. Bolles, “Random sample consensus: a paradigm for model ﬁtting with applications to image analysis and automated cartography,” Communications of the ACM, vol. 24, no. 6, pp. 381–395, 1981

work page 1981
[28]

Fine-tuning CNN image retrieval with no human annotation,

F. Radenovi ´c, G. Tolias, and O. Chum, “Fine-tuning CNN image retrieval with no human annotation,” TPAMI, 2018

work page 2018
[29]

Making large art historical photo archives searchable,

B. L. A. Seguin, “Making large art historical photo archives searchable,” EPFL, Tech. Rep., 2018

work page 2018

[1] [1]

New techniques for the digitization of art historical photographic archives-the case of the cini foundation in venice,

B. Seguin, L. Costiner, I. di Lenardo, and F. Kaplan, “New techniques for the digitization of art historical photographic archives-the case of the cini foundation in venice,” in Archiv- ing Conference , vol. 2018, no. 1. Society for Imaging Science and Technology, 2018, pp. 1–5

work page 2018

[2] [2]

Mas- sive 3d digitization of museum contents,

G. Guidi, S. G. Barsanti, L. L. Micoli, and M. Russo, “Mas- sive 3d digitization of museum contents,” in Built heritage: Monitoring conservation management . Springer, 2015, pp. 335–346

work page 2015

[3] [3]

Visual link retrieval in a database of paintings,

B. Seguin, C. Striolo, F. Kaplan et al., “Visual link retrieval in a database of paintings,” in European Conference on Computer Vision. Springer, 2016, pp. 753–767

work page 2016

[4] [4]

Tracking transmis- sion of details in paintings

B. Seguin, I. diLenardo, and F. Kaplan, “Tracking transmis- sion of details in paintings.” in DH, 2017

work page 2017

[5] [5]

The replica project: building a visual search engine for art historians,

B. Seguin, “The replica project: building a visual search engine for art historians,” XRDS: Crossroads, The ACM Magazine for Students , vol. 24, no. 3, pp. 24–29, 2018

work page 2018

[6] [6]

Artistic image classiﬁcation: An analysis on the printart database,

G. Carneiro, N. P. da Silva, A. Del Bue, and J. P. Costeira, “Artistic image classiﬁcation: An analysis on the printart database,” in European Conference on Computer Vision . Springer, 2012, pp. 143–157

work page 2012

[7] [7]

Classiﬁcation of artistic styles using binarized features derived from a deep neural network,

Y . Bar, N. Levy, and L. Wolf, “Classiﬁcation of artistic styles using binarized features derived from a deep neural network,” in Workshop at the European Conference on Computer Vision. Springer, 2014, pp. 71–84

work page 2014

[8] [8]

Large-scale Classification of Fine-Art Paintings: Learning The Right Metric on The Right Feature

B. Saleh and A. Elgammal, “Large-scale classiﬁcation of ﬁne- art paintings: Learning the right metric on the right feature,” arXiv preprint arXiv:1505.00855 , 2015

work page internal anchor Pith review Pith/arXiv arXiv 2015

[9] [9]

Towards a hierarchical multitask classiﬁcation framework for cultural heritage,

A. Belhi, A. Bouras, and S. Foufou, “Towards a hierarchical multitask classiﬁcation framework for cultural heritage,” in 2018 IEEE/ACS 15th International Conference on Computer Systems and Applications (AICCSA) . IEEE, 2018, pp. 1–7

work page 2018

[10] [10]

Detecting people in artwork with cnns,

N. Westlake, H. Cai, and P. Hall, “Detecting people in artwork with cnns,” in European Conference on Computer Vision . Springer, 2016, pp. 825–841

work page 2016

[11] [11]

Weakly supervised object detection in artworks,

N. Gonthier, Y . Gousseau, S. Ladjal, and O. Bonfait, “Weakly supervised object detection in artworks,” in European Con- ference on Computer Vision . Springer, 2018, pp. 692–709

work page 2018

[12] [12]

How to read paintings: Semantic art understanding with multi-modal retrieval,

N. Garcia and G. V ogiatzis, “How to read paintings: Semantic art understanding with multi-modal retrieval,” in European Conference on Computer Vision . Springer, 2018, pp. 676– 691

work page 2018

[13] [13]

The rijksmuseum challenge: Museum-centered visual recognition,

T. Mensink and J. Van Gemert, “The rijksmuseum challenge: Museum-centered visual recognition,” in Proceedings of In- ternational Conference on Multimedia Retrieval . ACM, 2014, p. 451

work page 2014

[14] [14]

Recognizing image style,

S. Karayev, M. Trentacoste, H. Han, A. Agarwala, T. Darrell, A. Hertzmann, and H. Winnemoeller, “Recognizing image style,” arXiv preprint arXiv:1311.3715 , 2013

work page arXiv 2013

[15] [15]

The state of the art: Object retrieval in paintings using discriminative regions

E. Crowley and A. Zisserman, “The state of the art: Object retrieval in paintings using discriminative regions.” in BMVC, 2014

work page 2014

[16] [16]

In search of art,

E. J. Crowley and A. Zisserman, “In search of art,” in Workshop at the European Conference on Computer Vision . Springer, 2014, pp. 54–70

work page 2014

[17] [17]

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

J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, “You only look once: Uniﬁed, real-time object detection,” in Pro- ceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 779–788

work page 2016

[18] [18]

The Cross-Depiction Problem: Computer Vision Algorithms for Recognising Objects in Artwork and in Photographs

H. Cai, Q. Wu, T. Corradi, and P. Hall, “The cross- depiction problem: Computer vision algorithms for recognis- ing objects in artwork and in photographs,” arXiv preprint arXiv:1505.00110, 2015

work page internal anchor Pith review Pith/arXiv arXiv 2015

[19] [19]

Data-driven visual similarity for cross-domain image match- ing,

A. Shrivastava, T. Malisiewicz, A. Gupta, and A. A. Efros, “Data-driven visual similarity for cross-domain image match- ing,” ACM Transactions on Graphics (ToG) , vol. 30, no. 6, p. 154, 2011

work page 2011

[20] [20]

Deep shape match- ing,

F. Radenovic, G. Tolias, and O. Chum, “Deep shape match- ing,” in ECCV, 2018

work page 2018

[21] [21]

Pose search: retrieving people using their pose,

V . Ferrari, M. Marin-Jimenez, and A. Zisserman, “Pose search: retrieving people using their pose,” in Computer Vision and Pattern Recognition, 2009. CVPR 2009. IEEE Conference on. IEEE, 2009, pp. 1–8

work page 2009

[22] [22]

2d articulated human pose estimation and retrieval in (almost) unconstrained still images,

M. Eichner, M. Marin-Jimenez, A. Zisserman, and V . Fer- rari, “2d articulated human pose estimation and retrieval in (almost) unconstrained still images,” International journal of computer vision, vol. 99, no. 2, pp. 190–214, 2012

work page 2012

[23] [23]

Visual sentences for pose retrieval over low-resolution cross-media dance collections,

R. Ren and J. Collomosse, “Visual sentences for pose retrieval over low-resolution cross-media dance collections,” IEEE Transactions on Multimedia , vol. 14, no. 6, pp. 1652–1661, 2012

work page 2012

[24] [24]

OpenPose: Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields

Z. Cao, G. Hidalgo, T. Simon, S.-E. Wei, and Y . Sheikh, “OpenPose: realtime multi-person 2D pose estimation using Part Afﬁnity Fields,” in arXiv preprint arXiv:1812.08008 , 2018

work page internal anchor Pith review Pith/arXiv arXiv 2018

[25] [25]

RMPE: Regional multi-person pose estimation,

H.-S. Fang, S. Xie, Y .-W. Tai, and C. Lu, “RMPE: Regional multi-person pose estimation,” in ICCV, 2017

work page 2017

[26] [26]

Mask r- cnn,

K. He, G. Gkioxari, P. Doll ´ar, and R. Girshick, “Mask r- cnn,” in Computer Vision (ICCV), 2017 IEEE International Conference on. IEEE, 2017, pp. 2980–2988

work page 2017

[27] [27]

Random sample consensus: a paradigm for model ﬁtting with applications to image analysis and automated cartography,

M. A. Fischler and R. C. Bolles, “Random sample consensus: a paradigm for model ﬁtting with applications to image analysis and automated cartography,” Communications of the ACM, vol. 24, no. 6, pp. 381–395, 1981

work page 1981

[28] [28]

Fine-tuning CNN image retrieval with no human annotation,

F. Radenovi ´c, G. Tolias, and O. Chum, “Fine-tuning CNN image retrieval with no human annotation,” TPAMI, 2018

work page 2018

[29] [29]

Making large art historical photo archives searchable,

B. L. A. Seguin, “Making large art historical photo archives searchable,” EPFL, Tech. Rep., 2018

work page 2018