Occlusion Robust Face Recognition Based on Mask Learning with PairwiseDifferential Siamese Network

Song, Lingxue; Gong, Dihong; Li, Zhifeng; Liu, Changsong; Liu, Wei

doi:10.48550/arxiv.1908.06290

Cited by 10 publications

(5 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Data Augmentation/ Recovery [15], [20], [21], [24], [26], [31], [35], [51], [54], [62], [64], [73], [78], [89], [90], [107], [112], [136], [140], [147], [183] Feature Extraction [1], [2], [11], [23], [27], [32], [33], [48], [49], [53], [60], [62], [65], [74], [80], [88], [93], [108], [109], [124], [125], [128], [134], [143], [148], [153], [162], [163], [169],…”

Section: Pipeline Category Publicationmentioning

confidence: 99%

“…MaskNet is defined as a shallow convolutional network, which is expected to assign lower weights to hidden units activated by the occluded facial areas. Recently, Song et al [139] propose a pairwise differential siamese network (PDSN) to build the correspondence between the occluded facial block and the corrupted feature elements with a mask learning strategy. The results show improved performance on synthesised and realistic occluded face datasets.…”

Section: Visible Part Selectionmentioning

confidence: 99%

“…Random patch [12] 2) StaƟsƟcal learning 3) Sparse representaƟon classifier 4) Deep learning 1) Subspace learning PCA [143], LDA [8] SOM [139] SRC [158], GRRC [169], RCSLBP [193] Data augmentaƟon [92], DDRC [11] F I G U R E 5 Categorisation of methods used in occlusion robust feature extraction approaches Unlike regular images, face images share common constraints, such as containing a smooth surface and regular texture. Face images are in fact confined to a face subspace.…”

Section: Learning-based Featuresmentioning

confidence: 99%

“…New occlusion detection techniques need to be designed to take advantage of massive public datasets. Better occlusion detection not only makes it possible to generate occlusion templates automatically but also provides an informative preprocessing for face recognition [138,139]. Most methods reviewed address the occlusion detection problem by partitioning the face into patches using a binary classifier, resulting in a rough occlusion area.…”

Section: Future Research Challengesmentioning

confidence: 99%

See 3 more Smart Citations

A survey of face recognition techniques under occlusion

2021

View full text Add to dashboard Cite

The limited capacity to recognise faces under occlusions is a long-standing problem that presents a unique challenge for face recognition systems and even humans. The problem regarding occlusion is less covered by research when compared to other challenges such as pose variation, different expressions, etc. Nevertheless, occluded face recognition is imperative to exploit the full potential of face recognition for real-world applications. In this article, the scope to occluded face recognition is restricted and a systematic categorisation that new as well as classic methods fit into is presented. First, the authors explore the kind of the occlusion problem and the type of inherent difficulties that can arise. As a part of this review, face detection under occlusion, a preliminary step in face recognition. Second the authors analyse how the existing face recognition methods cope with the occlusion problem and classify them into three categories, which are given as: 1) occlusion robust feature extraction approaches, 2) occlusion aware face recognition approaches, and 3) occlusion recovery based face recognition approaches. Furthermore, the motivations, innovations, pros and cons, and the performance of representative approaches for comparison are analyzed. Finally, future challenges and method trends of occluded face recognition are thoroughly discussed.This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

show abstract

Section: Pipeline Category Publicationmentioning

confidence: 99%

Section: Visible Part Selectionmentioning

confidence: 99%

Section: Learning-based Featuresmentioning

confidence: 99%

Section: Future Research Challengesmentioning

confidence: 99%

See 2 more Smart Citations

A survey of face recognition techniques under occlusion

2021

View full text Add to dashboard Cite

show abstract

“…The paper analyses how the proposed method helps in MFR and illustrates the basic principles of its anti-noise. Song et al (2019) proposed a mask learning method to repair damaged feature elements in human faces to develop the Pairwise Differential Siamese Network (PDSN). Masking dictionary using the difference between masked and unmasked face pairs of convolutional features.…”

Section: Introductionmentioning

confidence: 99%

Masked face recognition based on facenet and genetic algorithm

Cheng,

Hsiao,

Hong

et al. 2023

Int. J. Appl. Sci. Eng.

View full text Add to dashboard Cite

This study proposes a face recognition method based on FaceNet for face masks. FaceNet converts faces into 128-dimensional feature vectors using a network of deep learning models. The smaller the Euclidean distance between different face images, the closer the person is to the same person; conversely, the larger the Euclidean distance, the more different the person is. Wearing a mask affects some feature vector dimensions of FaceNet conversion, resulting in reduced recognition ability. This study uses a Genetic Algorithm (GA) to select and remove the feature vectors affected by mask wearing. The experimental results show that the validation rate value of 55 features removed by the GA is increased from 0.550 to 0.650 for the original unremoved features.

show abstract