An End-to-End System for Unconstrained Face Verification with Deep Convolutional Neural Networks

Chen, Jun-Cheng; Ranjan, Rajeev; Kumar, Amit; Chen, Ching‐Hui; Patel, Vishal M.; Chellappa, Rama

doi:10.1109/iccvw.2015.55

Cited by 76 publications

(51 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Note that, we obtain 99.7% on VR@FAR=0.1% using the 6K pair-matching scores of the standard protocol. 6 Results computed from the features publicly provided by the authors. [24] N N 0.805 0.604 Face-Aug-Pose-Syn [23] N N 0.886 0.725 Deep Multipose [1] N N 0.787 -Pose aware FR [22] N N 0.826 0.652 TPE [28] N N 0.871 0.766 All-In-One [25] N N 0.893 0.787…”

Section: Results and Evaluationmentioning

confidence: 99%

DeepVisage: Making Face Recognition Simple Yet With Powerful Generalization Skills

Hasnat

Bohné

Milgram

et al. 2017

2017 IEEE International Conference on Computer Vision Workshops (ICCVW)

View full text Add to dashboard Cite

Face recognition (FR) methods report significant performance by adopting the convolutional neural network (CNN) based learning methods. Although CNNs are mostly trained by optimizing the softmax loss, the recent trend shows an improvement of accuracy with different strategies, such as task-specific CNN learning with different loss functions, fine-tuning on target dataset, metric learning and concatenating features from multiple CNNs. Incorporating these tasks obviously requires additional efforts. Moreover, it demotivates the discovery of efficient CNN models for FR which are trained only with identity labels. We focus on this fact and propose an easily trainable and single CNN based FR method. Our CNN model exploits the residual learning framework. Additionally, it uses normalized features to compute the loss. Our extensive experiments show excellent generalization on different datasets. We obtain very competitive and state-of-the-art results on the LFW, IJB-A, YouTube faces and CACD datasets.

show abstract

Section: Results and Evaluationmentioning

confidence: 99%

DeepVisage: Making Face Recognition Simple Yet With Powerful Generalization Skills

Hasnat

Bohné

Milgram

et al. 2017

2017 IEEE International Conference on Computer Vision Workshops (ICCVW)

View full text Add to dashboard Cite

show abstract

“…For example, [24] aggregates local descriptors (RootSIFT [3]) extracted from face crops using Fisher Vector [26] (FV) encoding to obtain a single descriptor per face track. Since the success of deep learning in image-based face recognition [4,23,25,29,30,31,44], simple strategies for face descriptor aggregation prevailed, such as average-and max-pooling [7,25]. However, none of these strategies are trained end-to-end for face recognition as typically only the face descriptors are learnt, while aggregation is performed post hoc.…”

Section: Related Workmentioning

confidence: 99%

“…template descriptors of the same subject should be close to each other in the descriptor space, whereas those of different subjects should be far apart. Although common aggregation strategies, such as average pooling and max-pooling, are able to aggregate face descriptors to produce a compact template representation [5,7,25] and currently achieves the state-of-the-art results [5], we seek a better solution in this paper.…”

Section: Introductionmentioning

confidence: 99%

GhostVLAD for Set-Based Face Recognition

Zhong

Arandjelović

Zisserman

2019

Lecture Notes in Computer Science

View full text Add to dashboard Cite

The objective of this paper is to learn a compact representation of image sets for template-based face recognition. We make the following contributions: first, we propose a network architecture which aggregates and embeds the face descriptors produced by deep convolutional neural networks into a compact fixed-length representation. This compact representation requires minimal memory storage and enables efficient similarity computation. Second, we propose a novel GhostVLAD layer that includes ghost clusters, that do not contribute to the aggregation. We show that a quality weighting on the input faces emerges automatically such that informative images contribute more than those with low quality, and that the ghost clusters enhance the network's ability to deal with poor quality images. Third, we explore how input feature dimension, number of clusters and different training techniques affect the recognition performance. Given this analysis, we train a network that far exceeds the state-of-the-art on the IJB-B face recognition dataset. This is currently one of the most challenging public benchmarks, and we surpass the state-of-the-art on both the identification and verification protocols.

show abstract

“…Interestingly, the variance on the similarity scores does not decrease as template sizes increase, rather they stay largely the same even as the mean similarity increases. Figure 5 (right) shows the effect of template size on ver- (top) Template adaptation compared with CNN encoding with metric learning using triplet similarity embedding [4,6] or Joint Bayesian embedding [21,23]. (bottom) Template adaptation compared with CNN encoding and 2D alignment [5,4].…”

Section: Negative Set Studymentioning

confidence: 99%

Template Adaptation for Face Verification and Identification

Crosswhite

Byrne

Stauffer

et al. 2017

2017 12th IEEE International Conference on Automatic Face &Amp; Gesture Recognition (FG 2017)

View full text Add to dashboard Cite

Face recognition performance evaluation has traditionally focused on one-to-one verification, popularized by the Labeled Faces in the Wild dataset [1] for imagery and the YouTubeFaces dataset [2] for videos. In contrast, the newly released IJB-A face recognition dataset [3] unifies evaluation of one-to-many face identification with one-to-one face verification over templates, or sets of imagery and videos for a subject. In this paper, we study the problem of template adaptation, a form of transfer learning to the set of media in a template. Extensive performance evaluations on IJB-A show a surprising result, that perhaps the simplest method of template adaptation, combining deep convolutional network features with template specific linear SVMs, outperforms the state-of-the-art by a wide margin. We study the effects of template size, negative set construction and classifier fusion on performance, then compare template adaptation to convolutional networks with metric learning, 2D and 3D alignment. Our unexpected conclusion is that these other methods, when combined with template adaptation, all achieve nearly the same top performance on IJB-A for template-based face verification and identification.

show abstract

An End-to-End System for Unconstrained Face Verification with Deep Convolutional Neural Networks

Cited by 76 publications

References 28 publications

DeepVisage: Making Face Recognition Simple Yet With Powerful Generalization Skills

DeepVisage: Making Face Recognition Simple Yet With Powerful Generalization Skills

GhostVLAD for Set-Based Face Recognition

Template Adaptation for Face Verification and Identification

Contact Info

Product

Resources

About