Deep Metric Learning via Facility Location

Song, Hyun Oh; Jegelka, Stefanie; Rathod, Vivek; Murphy, Kevin

doi:10.1109/cvpr.2017.237

Cited by 262 publications

(208 citation statements)

References 22 publications

(60 reference statements)

Supporting

Mentioning

205

Contrasting

Order By: Relevance

“…Cars196 contains 16,185 images belonging to 196 classes of cars. In our experiments, we follow the settings in [3], taking the first 98 classes (8, [38].…”

Section: Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Deep Metric Learning With Density Adaptivity

Yao

Pan

et al. 2020

IEEE Trans. Multimedia

View full text Add to dashboard Cite

The problem of distance metric learning is mostly considered from the perspective of learning an embedding space, where the distances between pairs of examples are in correspondence with a similarity metric. With the rise and success of Convolutional Neural Networks (CNN), deep metric learning (DML) involves training a network to learn a nonlinear transformation to the embedding space. Existing DML approaches often express the supervision through maximizing inter-class distance and minimizing intra-class variation. However, the results can suffer from overfitting problem, especially when the training examples of each class are embedded together tightly and the density of each class is very high. In this paper, we integrate density, i.e., the measure of data concentration in the representation, into the optimization of DML frameworks to adaptively balance inter-class similarity and intra-class variation by training the architecture in an endto-end manner. Technically, the knowledge of density is employed as a regularizer, which is pluggable to any DML architecture with different objective functions such as contrastive loss, N-pair loss and triplet loss. Extensive experiments on three public datasets consistently demonstrate clear improvements by amending three types of embedding with the density adaptivity. More remarkably, our proposal increases Recall@1 from 67.95% to 77.62%, from 52.01% to 55.64% and from 68.20% to 70.56% on Cars196, CUB-200-2011 and Stanford Online Products dataset, respectively.

show abstract

“…Cars196 contains 16,185 images belonging to 196 classes of cars. In our experiments, we follow the settings in [3], taking the first 98 classes (8, [38].…”

Section: Methodsmentioning

confidence: 99%

“…(3) N-Pair [7] trains DML with N-pair loss. (4) Clustering [38] is a structured prediction based DML model which can be optimized with clustering quality metric. (5) Contrastive [19] uses contrastive loss for DML training.…”

Section: B Evaluation Metrics and Compared Methodsmentioning

confidence: 99%

Deep Metric Learning With Density Adaptivity

Yao

Pan

et al. 2020

IEEE Trans. Multimedia

View full text Add to dashboard Cite

show abstract

“…The superscript denotes the embedding size. In [24] Song et al claim the results in the N-pair [23] paper have been achieved by an average of ten extracted embeddings from ten random crops. The usage of such a crop averaging technique is marked with .…”

Section: Comparison To the State-of-the-artmentioning

confidence: 99%

“…Not all listed approaches employ the GoogLeNet architecture [26]. A ResNet50 v2 [8] with a top-1 accuracy of 75.6% on the ImageNet validation set [20] is used by Margin and InceptionBN [11] with 73.9% by Proxy-NCA [18] and Clustering [24]. Compared to the GoogLeNet, the two more advanced architectures might give a better general image retrieval performance.…”

Section: Comparison To the State-of-the-artmentioning

confidence: 99%

“…This is subject to future research. [25] GoogLeNet --62.1 Angular Loss 512 [28] GoogLeNet 54.7 71.4 70.9 Clustering 64 [24] InceptionBN 48.2 58.1 67.0 N-pair 64 [23] GoogLeNet 51.0 71.1 -N-pair 512 [23] GoogLeNet --67.7 PDDM+Quad. 128 [ V. CONCLUSION In this paper, we propose Nonlinear Rank Approximation loss (NRA) for deep metric learning, which significantly improves upon existing approaches like Triplet, Lifted Structured, and N-pair loss.…”

Section: Comparison To the State-of-the-artmentioning

confidence: 99%

See 1 more Smart Citation

Deep Metric Learning using Similarities from Nonlinear Rank Approximations

Schall

Barthel

Hezel

et al. 2019

2019 IEEE 21st International Workshop on Multimedia Signal Processing (MMSP)

View full text Add to dashboard Cite

In recent years, deep metric learning has achieved promising results in learning high dimensional semantic feature embeddings where the spatial relationships of the feature vectors match the visual similarities of the images. Similarity search for images is performed by determining the vectors with the smallest distances to a query vector. However, high retrieval quality does not depend on the actual distances of the feature vectors, but rather on the ranking order of the feature vectors from similar images. In this paper, we introduce a metric learning algorithm that focuses on identifying and modifying those feature vectors that most strongly affect the retrieval quality. We compute normalized approximated ranks and convert them to similarities by applying a nonlinear transfer function. These similarities are used in a newly proposed loss function that better contracts similar and disperses dissimilar samples. Experiments demonstrate significant improvement over existing deep feature embedding methods on the CUB-200-2011, Cars196, and Stanford Online Products data sets for all embedding sizes.

show abstract