Triplet Distillation For Deep Face Recognition

Feng, Yixiang; Wang, Huan; Hu, Haoji; Yu, Lu; Wang, Wei; Wang, Shiyan

doi:10.1109/icip40778.2020.9190651

Cited by 34 publications

(16 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In general, the main usage of knowledge distillation is to transfer the knowledge from several networks to another one (Hinton, Dean, and Vinyals 2014;Feng et al 2020;David et al 2016). Our work is closely related to the following very recent knowledge distillation work for recognition.…”

Section: Knowledge Distillationmentioning

confidence: 99%

Pose-Invariant Face Recognition via Adaptive Angular Distillation

Zhang

Chen

Yang

et al. 2022

AAAI

View full text Add to dashboard Cite

Pose-invariant face recognition is a practically useful but challenging task. This paper introduces a novel method to learn pose-invariant feature representation without normalizing profile faces to frontal ones or learning disentangled features. We first design a novel strategy to learn pose-invariant feature embeddings by distilling the angular knowledge of frontal faces extracted by teacher network to student network, which enables the handling of faces with large pose variations. In this way, the features of faces across variant poses can cluster compactly for the same person to create a pose-invariant face representation. Secondly, we propose a Pose-Adaptive Angular Distillation loss to mitigate the negative effect of uneven distribution of face poses in the training dataset to pay more attention to the samples with large pose variations. Extensive experiments on two challenging benchmarks (IJB-A and CFP-FP) show that our approach consistently outperforms the existing methods.

show abstract

Section: Knowledge Distillationmentioning

confidence: 99%

Pose-Invariant Face Recognition via Adaptive Angular Distillation

Zhang

Chen

Yang

et al. 2022

AAAI

View full text Add to dashboard Cite

show abstract

“…Previous works [14], [21] indicated that utilizing tripletbased learning is beneficial for learning discriminative face embeddings. Let x ∈ X represents a batch of training samples, and f (x) is the face embeddings obtained from the face recognition model.…”

Section: ) Self-restrained Triplet Lossmentioning

confidence: 99%

Self-restrained Triplet Loss for Accurate Masked Face Recognition

Damer¹,

Kirchbuchner²,

Kuijper³

2021

Preprint

View full text Add to dashboard Cite

Using the face as a biometric identity trait is motivated by the contactless nature of the capture process and the high accuracy of the recognition algorithms. After the current COVID-19 pandemic, wearing a face mask has been imposed in public places to keep the pandemic under control. However, face occlusion due to wearing a mask presents an emerging challenge for face recognition systems. In this paper, we presented a solution to improve the masked face recognition performance. Specifically, we propose the Embedding Unmasking Model (EUM) operated on top of existing face recognition models. We also propose a novel loss function, the Self-restrained Triplet (SRT), which enabled the EUM to produce embeddings similar to these of unmasked faces of the same identities. The achieved evaluation results on two face recognition models and two real masked datasets proved that our proposed approach significantly improves the performance in most experimental settings.

show abstract

“…Several effective face recognition models are trained using methods involving knowledge distillation [34]. In [35], an enhanced version of triplet loss is proposed, named triplet distillation, which exploits the capability of a teacher model to transfer the similarity information to a small model by adaptively varying the margin between positive and negative pairs. In [36], the authors present a novel model compres-sion approach based on the student-teacher paradigm for face recognition applications, which consists of a training teacher network at greater image resolution while student networks are trained at lower image resolutions than that of teacher network.…”

Section: Related Workmentioning

confidence: 99%

Face Recognition Based on Lightweight Convolutional Neural Networks

2021

View full text Add to dashboard Cite

Face recognition algorithms based on deep learning methods have become increasingly popular. Most of these are based on highly precise but complex convolutional neural networks (CNNs), which require significant computing resources and storage, and are difficult to deploy on mobile devices or embedded terminals. In this paper, we propose several methods to improve the algorithms for face recognition based on a lightweight CNN, which is further optimized in terms of the network architecture and training pattern on the basis of MobileFaceNet. Regarding the network architecture, we introduce the Squeeze-and-Excitation (SE) block and propose three improved structures via a channel attention mechanism—the depthwise SE module, the depthwise separable SE module, and the linear SE module—which are able to learn the correlation of information between channels and assign them different weights. In addition, a novel training method for the face recognition task combined with an additive angular margin loss function is proposed that performs the compression and knowledge transfer of the deep network for face recognition. Finally, we obtained high-precision and lightweight face recognition models with fewer parameters and calculations that are more suitable for applications. Through extensive experiments and analysis, we demonstrate the effectiveness of the proposed methods.

show abstract

Triplet Distillation For Deep Face Recognition

Cited by 34 publications

References 15 publications

Pose-Invariant Face Recognition via Adaptive Angular Distillation

Pose-Invariant Face Recognition via Adaptive Angular Distillation

Self-restrained Triplet Loss for Accurate Masked Face Recognition

Face Recognition Based on Lightweight Convolutional Neural Networks

Contact Info

Product

Resources

About