Pith. sign in

REVIEW 3 major objections 6 minor 38 references

Metric Learning With HORDE: High-Order Regularizer for Deep Embeddings

T0 review · 3 major / 6 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 Matching high-order moments of deep-feature maps, not just their mean, reduces the scattering problem and outperforms previous metric-learning methods on four image-retrieval benchmarks.

desk verdict Solid empirical regularizer for deep metric learning; the theory section is not proven, but consistent gains and released code carry the paper. read the letter →

arxiv 1908.02735 v1 pith:BG6Y5MRU submitted 2019-08-07 cs.CV

Pierre Jacob , David Picard , Aymeric Histace , Edouard Klein This is my paper · ORCID
classification cs.CV
keywords deepmetriclearninghigh-ordermomentsscatteringproblemMaximumMeanDiscrepancyWassersteindistanceRandomMaclaurinfactorizationimageretrievalregularization
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

Deep metric-learning models usually represent an image by the average of its deep convolutional features, and that average is trained to be discriminative. This paper argues that the average ignores the shape of the deep-feature distribution, which tends to be scattered across feature space, making the representation fragile to occlusion, cropping, and background variation. HORDE, the proposed regularizer, enforces that visually similar images have deep features drawn from nearly the same distribution and dissimilar images from non-overlapping distributions by matching or separating high-order moments, which are spread-and-shape statistics such as variance and skewness. The paper proves that the distance between these moments upper-bounds the Maximum Mean Discrepancy and lower-bounds the Wasserstein distance between feature distributions, so moment matching is a justified proxy for alignment. On CUB-200-2011, Cars-196, Stanford Online Products, and In-Shop Clothes Retrieval, the regularizer improves several standard losses and an ensemble method, giving state-of-the-art retrieval accuracy.

What carries the argument

The load-bearing object is the approximated high-order moment feature map, built with a Random Maclaurin factorization: $\varphi_k(x) = (W_1^\top x) \odot (W_2^\top x) \odot \cdots \odot (W_k^\top x)$, computed recursively so that the k-th moment reuses the previous one. Each moment is pooled over the image, embedded, and fed into a metric-learning loss, and the HORDE loss is the sum of these per-moment losses. The theoretical link is the Taylor expansion of the Gaussian kernel, which expresses the Maximum Mean Discrepancy as a weighted sum of differences of vectorized moments; this yields HORDE as an upper bound on MMD and, via the known relation between MMD and the Wasserstein distance, as a lower bound on the Wasserstein distance. That chain of inequalities is what makes optimizing moment distances a principled proxy for aligning the whole deep-feature distribution.

What would settle it

On CUB-200-2011, take a trained HORDE model and its baseline and measure the empirical Maximum Mean Discrepancy and Wasserstein distance between deep-feature sets of same-class image pairs; if HORDE reduces the moment differences without reducing these distribution-level distances, the claimed concentration mechanism is not what produces the retrieval improvements.

Watch

Extended reading notes

Core claim

The central claim is that the scattering problem, where deep features are spread out even when their mean is discriminative, can be corrected by regularizing the shape of the feature distribution rather than only its center. HORDE computes approximations of the second through K-th moments of the spatial deep features of each image, applies the same metric-learning loss to an embedding of each moment, and adds those per-moment losses to the standard loss. Minimizing moment distances for similar images drives their feature distributions to coincide, while maximizing them for dissimilar images keeps the distributions apart. The paper proves that the sum of moment differences is an upper bound on the squared Maximum Mean Discrepancy and a lower bound on the squared Wasserstein distance, so the regularizer is a tractable stand-in for full distribution matching. Across four benchmarks, HORDE consistently improves contrastive, triplet, and binomial-deviance baselines, and it also raises the performance of the ABE ensemble method, producing state-of-the-art Recall@1 results.

Load-bearing premise

The regularizer treats the spatial features of a feature map as independent samples from a per-image distribution, so the empirical mean of each moment must be a faithful estimate; if those features are strongly correlated or position-dependent, matching moments may not align the true distributions.

Editorial extensions

If this is right

  • On CUB-200-2011 and Cars-196, adding HORDE to contrastive, triplet, or binomial-deviance losses improves Recall@1 compared with the same backbone without the regularizer.
  • HORDE also improves the ABE ensemble method, reaching state-of-the-art Recall@1 of 62.7% on CUB-200-2011 and 86.4% on Cars-196.
  • Concatenating the learned high-order moments at test time, then reducing them with PCA to the embedding dimension, gives further retrieval gains over using only the mean representation.
  • The regularizer transfers to large-scale retrieval, raising the binomial-deviance baseline by 5.2% Recall@1 on Stanford Online Products and by 3.1% on In-Shop Clothes Retrieval.
  • Training with five or six moments is reported to be more stable and to reach peak accuracy about one thousand steps sooner than the baseline on CUB-200-2011.

Reading between the lines

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

  • If moment matching is what actually concentrates deep-feature distributions, the same regularizer could help other tasks that aggregate deep features, such as face verification or person re-identification, beyond the four retrieval benchmarks tested.
  • The independent-samples assumption suggests a testable extension: weighting spatial features by a smoothness kernel, so neighboring positions count less redundantly, might make the moment estimates more faithful when features are spatially correlated.
  • Because the theory ties moment matching to MMD and Wasserstein distances, HORDE could plausibly serve as a cheap distribution-alignment regularizer in domain adaptation or generative modeling, where matching feature distributions across domains is the explicit goal.
  • The stability gains reported with higher moment orders could be explored further by scheduling moment orders during training, for instance adding them one at a time, to separate a curriculum effect from pure regularization strength.
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.

Referee Report

3 major / 6 minor

Summary. The paper proposes HORDE, a regularizer for deep metric learning that appends DML-style losses on high-order moment approximations of the spatial deep-feature distribution. The moments are computed with Random Maclaurin features, with a recursive/cascaded variant whose projection matrices can be learned. The authors claim that pulling moments of similar-image features together and pushing apart those of dissimilar images reduces the "scattering" of deep features and makes the average-pooled representation more robust. They state two theorems: that the high-order moment distance upper-bounds the squared MMD and lower-bounds the squared 1-Wasserstein distance. They report Recall@K improvements over several baselines on CUB-200-2011, Cars-196, Stanford Online Products, and In-Shop Clothes Retrieval, including an ensemble variant, and provide ablations on CUB.

Significance. The distribution-aware regularization idea is simple and well motivated, and the empirical contribution is credible: the paper reports consistent gains across four datasets, multiple loss functions, and an ensemble method, with code released and ablations that isolate the Random Maclaurin variant, the learned projections, and the cascaded architecture. If the theoretical claims were correct, the paper would provide a clean connection between moment matching and distributional distances. As it stands, however, the proofs of both theorems contain load-bearing gaps, and the implemented objective is not obviously the quantity analyzed in the theorems. The empirical value of the method is therefore the main basis for publication, and the theoretical section must be corrected or substantially weakened before the paper can be accepted.

major comments (3)
  1. [Section 3.2, Theorem 2 (Eqs. (18)-(21))] The claimed positive lower bound is not established. In Eq. (19), a' is defined as min_k a_k with a_k = (2γ)^k/k!; since a_k tends to 0 as k tends to infinity, a' = 0 and therefore a = αa'/K = 0 in Eq. (21), making the bound vacuous. If the minimum is intended only over 1 ≤ k ≤ p, the proof still does not justify the step from the pointwise kernel lower bound k(x,y) ≥ α∑ a_k <x^{⊗k}, y^{⊗k}> to an MMD lower bound. Because MMD²(I,J) = E[k(x,x')] + E[k(y,y')] - 2E[k(x,y)], a pointwise lower bound on k makes the negative cross term more negative while making the positive diagonal terms larger, so no inequality between the MMD values follows. The inference is invalid in general; for example, with the constant kernel k ≡ 1 and L(x,y) = xy on {±1}, L ≤ k pointwise but MMD_L²(δ_1,δ_{-1}) = 4 > MMD_k² = 0. Thus Theorem 2's Wasserstein lower bound is unsupported.
  2. [Section 3.2, Theorem 1 (Eqs. (13)-(17))] The same pointwise-to-MMD inference flaw invalidates the upper-bound proof. The proof uses Eq. (15), k(x,y) ≤ 1 + ∑_{k=1}^∞ a_k <x^{⊗k}, y^{⊗k}>, and then Eq. (17) to conclude an inequality on MMD². A pointwise bound between kernels does not imply the corresponding bound between MMD values, and the authors do not show that the difference kernel is positive definite, which would be the required condition. In addition, the constant '1' appearing in Eqs. (13) and (17) cannot arise from the feature map in Eq. (16): a constant feature contributes zero to the MMD of two probability measures, so the stated bound with the '+1' term does not follow from the proposed feature map even if the kernel domination were valid. Theorem 1 is therefore not proved as written.
  3. [Section 3.1 vs. 3.2, Eq. (8) and Algorithm 1] The regularizer actually optimized is not the quantity treated in Theorems 1-2. Equations (7)-(8) and Algorithm 1 apply a DML loss to random or learned Maclaurin projections of empirical spatial averages, whereas the theorems concern exact high-order moment tensors E_{x∼I}[x^{⊗k}] under a Gaussian kernel on R^c. No error bound or equivalence argument connects the projected empirical moments to the exact moments, and no argument shows that the learned projections preserve the moment-distance ordering used in the theorems. Furthermore, Section 3.2 models an image as a distribution from which spatial deep features are sampled i.i.d.; for convolutional feature maps, spatial positions are correlated and non-exchangeable, so this assumption is unverified and may not hold. The theoretical justification should either be repaired to cover the actual objective or removed from the central claims.
minor comments (6)
  1. [Section 3.2, Eq. (11)] The symbol K denotes both the number of high-order moments (Eq. (7)) and the uniform bound on moment norms (Eq. (11)); please use distinct symbols, for example M for the bound.
  2. [Section 3.2, Eq. (17)] In the second sum, the expectation is written as Ex∼J[x^{⊗k}], but it should be Ey∼J[y^{⊗k}].
  3. [Algorithm 1] The pseudocode fixes the projection matrices W1,...,WK as random {−1,+1} matrices, while the main text and Section 5.2 describe a trainable cascade; please state that Algorithm 1 describes the untrained variant or add the learning rule for the projectors.
  4. [Section 4] The sentence 'we train HORDE using 5 high-order moments with 5 classes and 8 images per instance per batch' is unclear; presumably it should read '8 images per class per batch'.
  5. [Tables 3-5] The k/n column layout is very hard to parse; grouping columns by k with explicit n labels, or splitting into subtables, would make the ablation results much clearer.
  6. [Theorems 1-2] The remainder terms o(γ^pK/p!) make the statements asymptotic in p, but the theorems are then invoked as if they justify arbitrary finite p; please state explicitly the range of p for which the bounds are intended.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: HORDE is evaluated on external benchmarks and the theoretical bounds are derived from independent definitions, not from the method's own fitted values.

full rationale

The central derivation chain is a regularizer defined by moment distances (Eq. 7-8), a theoretical connection through Taylor expansion of the Gaussian kernel (Eq. 15-21), and benchmark evaluation (Tables 1-2). None of these steps reduces to its inputs by construction. Hyperparameters (number of moments K, dimension 8192, embeddings 512) are fixed by the authors and not fitted to test labels; the test-time concatenation and PCA is a representation choice, not a label-based prediction. The theoretical analysis may contain correctness gaps, such as the potentially vanishing constant in Theorem 2, but a vacuous or invalid bound is a mathematical error rather than circular reasoning. Self-citations are limited to contextual references and are not load-bearing. The empirical improvements are measured on standard external benchmarks against independent baselines, so the derivation is self-contained.

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

The central claim rests on the i.i.d. sampling assumption for deep features and the bounded-moment assumption. The Random Maclaurin consistency is standard, but the trainable cascade is an extension. No new physical entities are introduced.

free parameters (4)
  • Number of high-order moments K = 5 (CUB/CARS), 4 (SOP/INSHOP)
    Selected per dataset; ablation shows performance and training stability depend on K.
  • Random projection dimension d = 8192
    Chosen for all high-order moment approximations; affects memory and approximation quality.
  • Embedding dimension = 512
    Standard embedding size used for all moments and the image representation.
  • Batch composition = 5 classes x 8 images (CUB/CARS); 2 images x 40 classes (SOP/INSHOP)
    Chosen to accommodate dataset class sizes; affects the quality of moment estimates.
assumptions (4)
  • domain assumption Deep features within an image are treated as i.i.d. samples from an image-specific distribution.
    Section 3.2 states 'An image is then represented as a distribution I from which we can sample a set of deep features'. The method relies on empirical moments of these samples being meaningful.
  • domain assumption All moments of the feature distributions exist and are bounded by a common constant K (Eq. 11).
    Used in Theorems 1 and 2 to justify truncating the Taylor expansion; not verified for CNN features.
  • standard math The Random Maclaurin approximation provides a consistent estimator of the inner product between high-order moments (Eq. 2).
    Standard result from Kar and Karnick (2012), cited as [10]. In practice the projectors are trained in the cascade, which goes beyond the original consistency guarantee.
  • standard math MMD is a lower bound of the Wasserstein distance (Eq. 20).
    Known result cited from Sriperumbudur et al. (2010), reference [24].

how reviews work

0 comments
Cite this review

Pith. "Pith review of Metric Learning With HORDE: High-Order Regularizer for Deep Embeddings." pith.science (2026). https://pith.science/paper/BG6Y5MRU

@misc{pith2026190802735,
  author       = {Pith},
  title        = {Pith review of: Metric Learning With HORDE: High-Order Regularizer for Deep Embeddings},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BG6Y5MRU}},
  note         = {Machine review of arXiv:1908.02735}
}
read the original abstract

Learning an effective similarity measure between image representations is key to the success of recent advances in visual search tasks (e.g. verification or zero-shot learning). Although the metric learning part is well addressed, this metric is usually computed over the average of the extracted deep features. This representation is then trained to be discriminative. However, these deep features tend to be scattered across the feature space. Consequently, the representations are not robust to outliers, object occlusions, background variations, etc. In this paper, we tackle this scattering problem with a distribution-aware regularization named HORDE. This regularizer enforces visually-close images to have deep features with the same distribution which are well localized in the feature space. We provide a theoretical analysis supporting this regularization effect. We also show the effectiveness of our approach by obtaining state-of-the-art results on 4 well-known datasets (Cub-200-2011, Cars-196, Stanford Online Products and Inshop Clothes Retrieval).

Figures

Figures reproduced from arXiv: 1908.02735 by the authors.

Figure 1
Figure 1. 2D visualizations of representations (stars) and deep features (points) using t-SNE computed from a DML architec [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Global overview of our HORDE architecture. The deep convolutional neural network extracts h × w × c deep features. The standard architecture (top blue block) relies on a global average pooling and an embedding before computing the LDML loss. The bottom red block is our HORDE regularizer, composed by the approximation of all high-order moments φk, global average pooling and embeddings before computing the sum of each… view at source ↗
Figure 3
Figure 3. Qualitative results on CUB for HORDE. Correct results are highlighted green (incorrect in red). ing the high-order moments as representations, we observe improved results with respect to the baseline for orders 2 and 3. Note however that the reported high-order results are not comparable to the first order as the similarity measure is computed on the 8192 dimensional representations. While adding orders higher than … view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

38 extracted references · 38 canonical work pages

  1. [1]

    Cross- modal retrieval in the cooking context: Learning se- mantic text-image embeddings

    Micael Carvalho, R ´emi Cad`ene, David Picard, Laure Soulier, Nicolas Thome, and Matthieu Cord. Cross- modal retrieval in the cooking context: Learning se- mantic text-image embeddings. In The 41st Interna- tional ACM SIGIR Conference on Research & Devel- opment in Information Retrieval, 2018. 1

    work page 2018

  2. [2]

    Beyond triplet loss: A deep quadru- plet network for person re-identification

    Weihua Chen, Xiaotang Chen, Jianguo Zhang, and Kaiqi Huang. Beyond triplet loss: A deep quadru- plet network for person re-identification. In The IEEE Conference on Computer Vision and Pattern Recogni- tion (CVPR), July 2017. 2

    work page 2017

  3. [3]

    Learn- ing a similarity metric discriminatively, with appli- cation to face verification

    Sumit Chopra, Raia Hadsell, and Yann LeCun. Learn- ing a similarity metric discriminatively, with appli- cation to face verification. In The IEEE Conference on Computer Vision and Pattern Recognition (CVPR),

    work page

  4. [4]

    Deep adversarial metric learning

    Yueqi Duan, Wenzhao Zheng, Xudong Lin, Jiwen Lu, and Jie Zhou. Deep adversarial metric learning. In The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2018. 2

    work page 2018

  5. [5]

    Deep metric learning with hierarchical triplet loss

    Weifeng Ge. Deep metric learning with hierarchical triplet loss. In The European Conference on Computer Vision (ECCV), September 2018. 6, 7

    work page 2018

  6. [6]

    A kernel method for the two-sample-problem

    Arthur Gretton, Karsten M Borgwardt, Malte Rasch, Bernhard Sch ¨olkopf, and Alex J Smola. A kernel method for the two-sample-problem. In Advances in neural information processing systems, pages 513– 520, 2007. 4, 5

    work page 2007

  7. [7]

    Smart mining for deep metric learning

    Ben Harwood, Vijay Kumar B G, Gustavo Carneiro, Ian Reid, and Tom Drummond. Smart mining for deep metric learning. In The IEEE International Confer- ence on Computer Vision (ICCV), Oct 2017. 2

    work page 2017

  8. [8]

    Batch normaliza- tion: Accelerating deep network training by reducing internal covariate shift

    Sergey Ioffe and Christian Szegedy. Batch normaliza- tion: Accelerating deep network training by reducing internal covariate shift. In Proceedings of the 32nd International Conference on Machine Learning , Jul

    work page

Show all 38 references
  1. [9]

    Negative evidences and co-occurrences in image retrieval: the benefit of PCA and whitening

    Herv ´e J ´egou and Ondrej Chum. Negative evidences and co-occurrences in image retrieval: the benefit of PCA and whitening. In The European Conference on Computer Vision (ECCV), Oct. 2012. 3

    2012

  2. [10]

    Random fea- ture maps for dot product kernels

    Purushottam Kar and Harish Karnick. Random fea- ture maps for dot product kernels. In Proceedings of the Fifteenth International Conference on Artificial In- telligence and Statistics, Apr 2012. 3

    2012

  3. [11]

    Attention-based ensemble for deep metric learning

    Wonsik Kim, Bhavya Goyal, Kunal Chawla, Jungmin Lee, and Keunjoo Kwon. Attention-based ensemble for deep metric learning. In The European Conference on Computer Vision (ECCV), September 2018. 2, 6, 7

    2018

  4. [12]

    3d object representations for fine-grained cate- gorization

    Jonathan Krause, Michael Stark, Jia Deng, and Li Fei- Fei. 3d object representations for fine-grained cate- gorization. In 4th International IEEE Workshop on 3D Representation and Recognition (3dRR-13) , Dec

  5. [13]

    Deep variational metric learning

    Xudong Lin, Yueqi Duan, Qiyuan Dong, Jiwen Lu, and Jie Zhou. Deep variational metric learning. InThe European Conference on Computer Vision (ECCV) , September 2018. 2, 3, 6

    2018

  6. [14]

    Deep relative distance learning: Tell the difference between similar vehicles

    Hongye Liu, Yonghong Tian, Yaowei Yang, Lu Pang, and Tiejun Huang. Deep relative distance learning: Tell the difference between similar vehicles. In The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2016. 1

    2016

  7. [15]

    Deepfashion: Powering robust clothes recognition and retrieval with rich annotations

    Ziwei Liu, Ping Luo, Shi Qiu, Xiaogang Wang, and Xiaoou Tang. Deepfashion: Powering robust clothes recognition and retrieval with rich annotations. InPro- ceedings of IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2016. 1, 2, 6

    2016

  8. [16]

    Leung, Sergey Ioffe, and Saurabh Singh

    Yair Movshovitz-Attias, Alexander Toshev, Thomas K. Leung, Sergey Ioffe, and Saurabh Singh. No fuss distance metric learning using proxies. In The IEEE International Conference on Computer Vision (ICCV), Oct 2017. 2

    2017

  9. [17]

    Deep metric learning via facility lo- cation

    Hyun Oh Song, Stefanie Jegelka, Vivek Rathod, and Kevin Murphy. Deep metric learning via facility lo- cation. In The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), July 2017. 2

    2017

  10. [18]

    Deep metric learning via lifted struc- tured feature embedding

    Hyun Oh Song, Yu Xiang, Stefanie Jegelka, and Sil- vio Savarese. Deep metric learning via lifted struc- tured feature embedding. In The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2016. 1, 2, 6

    2016

  11. [19]

    Bier - boosting independent embed- dings robustly

    Michael Opitz, Georg Waltner, Horst Possegger, and Horst Bischof. Bier - boosting independent embed- dings robustly. In The IEEE International Conference on Computer Vision (ICCV), Oct 2017. 1, 2, 3, 6

    2017

  12. [20]

    Deep metric learning with BIER: boosting independent embeddings robustly

    Michael Opitz, Georg Waltner, Horst Possegger, and Horst Bischof. Deep metric learning with BIER: boosting independent embeddings robustly. IEEE transactions on pattern analysis and machine intelli- gence, 2018. 6

    2018

  13. [21]

    Metric learning with adaptive density discrimination

    Oren Rippel, Manohar Paluri, Piotr Dollar, and Lubomir Bourdev. Metric learning with adaptive density discrimination. International Conference on Learning Representations (ICLR), May 2016. 2

    2016

  14. [22]

    Facenet: A unified embedding for face recog- nition and clustering

    Florian Schroff, Dmitry Kalenichenko, and James Philbin. Facenet: A unified embedding for face recog- nition and clustering. In The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , June 2015. 1, 2

    2015

  15. [23]

    Improved deep metric learning with multi-class n-pair loss objective

    Kihyuk Sohn. Improved deep metric learning with multi-class n-pair loss objective. In Advances in Neu- ral Information Processing Systems 29, Dec 2016. 1, 2

    2016

  16. [24]

    Sriperumbudur, Arthur Gretton, Kenji Fukumizu, Bernhard Sch¨olkopf, and Gert R.G

    Bharath K. Sriperumbudur, Arthur Gretton, Kenji Fukumizu, Bernhard Sch¨olkopf, and Gert R.G. Lanck- riet. Hilbert space embeddings and metrics on proba- bility measures. J. Mach. Learn. Res., Aug 2010. 5

    2010

  17. [25]

    Going deeper with convolutions

    Christian Szegedy, Wei Liu, Yangqing Jia, Pierre Ser- manet, Scott Reed, Dragomir Anguelov, Dumitru Er- han, Vincent Vanhoucke, and Andrew Rabinovich. Going deeper with convolutions. In The IEEE Con- ference on Computer Vision and Pattern Recognition (CVPR), June 2015. 1, 6

    2015

  18. [26]

    Learning deep embeddings with histogram loss

    Evgeniya Ustinova and Victor Lempitsky. Learning deep embeddings with histogram loss. In Advances in Neural Information Processing Systems 29, Dec 2016. 1, 2, 7

    2016

  19. [27]

    The Caltech-UCSD Birds-200-2011 Dataset

    Catherine Wah, Steve Branson, Peter Welinder, Pietro Perona, and Serge Belongie. The Caltech-UCSD Birds-200-2011 Dataset. Technical Report CNS-TR- 2011-001, California Institute of Technology, 2011. 2, 6, 7

    2011

  20. [28]

    Cosface: Large margin cosine loss for deep face recognition

    Hao Wang, Yitong Wang, Zheng Zhou, Xing Ji, Di- hong Gong, Jingchao Zhou, Zhifeng Li, and Wei Liu. Cosface: Large margin cosine loss for deep face recognition. In The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2018. 2

    2018

  21. [29]

    Deep metric learning with angular loss

    Jian Wang, Feng Zhou, Shilei Wen, Xiao Liu, and Yuanqing Lin. Deep metric learning with angular loss. In The IEEE International Conference on Computer Vision (ICCV), Oct 2017. 2, 6

    2017

  22. [30]

    Xun Wang, Xintong Han, Weilin Huang, Dengke Dong, and Matthew R. Scott. Multi-similarity loss with general pair weighting for deep metric learning. In The IEEE Conference on Computer Vision and Pat- tern Recognition (CVPR), June 2019. 6, 7

    2019

  23. [31]

    Jnatas Wehrmann and Rodrigo C. Barros. Bidirec- tional retrieval made simple. In The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2018. 1

    2018

  24. [32]

    Deep asymmetric metric learning via rich re- lationship mining

    Xinyi Xu, Yanhua Yang, Cheng Deng, and Feng Zheng. Deep asymmetric metric learning via rich re- lationship mining. In The IEEE Conference on Com- puter Vision and Pattern Recognition (CVPR) , pages 4076 – 4085, June 2019. 6

    2019

  25. [33]

    Deep randomized ensembles for metric learning

    Hong Xuan, Richard Souvenir, and Robert Pless. Deep randomized ensembles for metric learning. In The European Conference on Computer Vision (ECCV), September 2018. 2

    2018

  26. [34]

    Correcting the triplet selection bias for triplet loss

    Baosheng Yu, Tongliang Liu, Mingming Gong, Changxing Ding, and Dacheng Tao. Correcting the triplet selection bias for triplet loss. In The European Conference on Computer Vision (ECCV) , September

  27. [35]

    Hard- aware deeply cascaded embedding

    Yuhui Yuan, Kuiyuan Yang, and Chao Zhang. Hard- aware deeply cascaded embedding. InThe IEEE Inter- national Conference on Computer Vision (ICCV), Oct

  28. [36]

    Hardness-aware deep metric learning

    Wenzhao Zheng, Zhaodong Chen, Jiwen Lu, and Jie Zhou. Hardness-aware deep metric learning. In The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 72 – 81, June 2019. 6

    2019

  29. [37]

    Learning deep features for discriminative localization

    Bolei Zhou, Aditya Khosla, Agata Lapedriza, Aude Oliva, and Antonio Torralba. Learning deep features for discriminative localization. In The IEEE Con- ference on Computer Vision and Pattern Recognition (CVPR), June 2016. 1, 2

    2016

  30. [38]

    Effi- cient online local metric adaptation via negative sam- ples for person re-identification

    Jiahuan Zhou, Pei Yu, Wei Tang, and Ying Wu. Effi- cient online local metric adaptation via negative sam- ples for person re-identification. In The IEEE Inter- national Conference on Computer Vision (ICCV), Oct

Pith tools

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