A Discriminative Feature Learning Approach for Deep Face Recognition

By Yandong Wen, Kaipeng Zhang, Zhifeng Li, Yu Qiao. 2016

4,029 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Representation & Self-Supervised Learning, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Representation & Self-Supervised Learning

In Signal Processing

Citations in the corpus

Unsupervised Embedding Learning via Invariant and Spreading Instance Feature (2019, background): “Deep embedding learning usually learns an embedding function by minimizing the intra-class variation and maximizing the inter-class variation [32, 37, 45, 47].”
Prototypical Networks for Few-shot Learning (2017, neutral): “[35] for face recognition is similar to ours but has two main differences.”
The Non-IID Data Quagmire of Decentralized Machine Learning (2019, methodology): “• F ACE R ECOGNITION with the center-loss face model (Wen et al., 2016) over the CASIA-WebFace (Yi et al., 2014) dataset.”
Exploring the Encoding Layer and Loss Function in End-to-End Speaker and Language Recognition System (2018, background): “With proper λ , the discriminative power of deep features can be signiﬁcantly enhanced [34].”
MultiGrain: a unified image embedding for classes and instances (2019, background): “, 2018) or using embedding loss at the class level (Wen et al., 2016) can also improve the accuracy.”
Large-Scale Long-Tailed Recognition in an Open World (2019, background): “Compared to alternatives [56, 48], this memory is appealing for our OLTR task: It is almost effortlessly and jointly learned alongside the direct features {vdirect n }, and it considers both intra-class compactness and inter-class discriminativeness.”
CosFace: Large Margin Cosine Loss for Deep Face Recognition (2018, methodology): “For [42], it only explicitly minimizes the intra-class variance while ignoring the inter-class variances, which may result in suboptimal solutions.”
Continual Representation Learning for Biometric Identification (2020, methodology): “ CL methods can improve the representation in the new setting, and our method achieves better results than the competitors. 1. Introduction Biometric identiﬁcation [23,51], including face recognition [11,37,67] and person re-identiﬁcation (re-id) [34,66, 71], has achieved signiﬁcant progress in the ”
A Survey of Convolutional Neural Networks: Analysis, Applications, and Prospects (2020, methodology): “Another one is the center loss [77], which is an improvement based upon cross entropy.”
Range Loss for Deep Face Recognition with Long-Tailed Training Data (2016, methodology): “More recently, [31] proposes center loss which takes account of class-clusters in CNN training.”
SphereFace Revived: Unifying Hyperspherical Face Recognition (2021, methodology): “Inspired by linear discriminant analysis, [14] propose center loss for CNNs and obtain promising performance.”
Central Similarity Quantization for Efficient Image and Video Retrieval (2019, methodology): “However, we empirically ﬁnd that learned centers by some common meth-ods [11, 43, 29] cannot provide better hash functions than the analytically constructed ones.”
ShrinkTeaNet: Million-scale Lightweight Face Recognition via Shrinking Teacher-Student Networks (2019, methodology): “Method Training Data Accuracy Center Loss [35] 0.”
Dense Relation Network: Learning Consistent and Context-Aware Representation for Semantic Image Segmentation (2018, methodology): “Therefore, inspired by Center Loss [24], we compose context-restricted loss for hierarchical contextual representations as below:”
Label Mapping Neural Networks with Response Consolidation for Class Incremental Learning (2019, background): “The reason behind this problem is that the traditional softmax layer suffers from high intra-class variations [14].”
Regular Polytope Networks (2021, methodology): “1) MNIST and CIFAR: We trained the so-called LeNet++ architecture [47] on all the MNIST family data sets.”
Circle Loss: A Unified Perspective of Pair Similarity Optimization (2020, methodology): “The former employs a classiﬁcation loss function ( e.g. , Softmax plus cross-entropy loss [25, 16, 36]) to optimize the similarity between samples and weight vectors.”
Face Recognition Systems: A Survey (2020, background): “[113] introduce a new supervision signal, called center loss, for the face recognition task in order to improve the discriminative power of the deeply learned features.”