Weighted Softmax Loss for Face Recognition via Cosine Distance

2018 ◽  
pp. 340-348
Author(s):  
Hu Zhang ◽  
Xianliang Wang ◽  
Zhixiang He
Keyword(s):  
Face Recognition ◽  
Cosine Distance

scholarly journals HUMAN IDENTIFICATION BASED ON FACE RECOGNITION SYSTEM

2021 ◽  
Vol 25 (01) ◽  
pp. 80-91
Author(s):  
Saba K. Naji ◽  
◽  
Muthana H. Hamd ◽  
Keyword(s):  
Face Recognition ◽  
Euclidean Distance ◽  
Local Binary Pattern ◽  
Texture Features ◽  
Recognition System ◽  
Manhattan Distance ◽  
Distance Measurements ◽  
Local Ternary Pattern ◽  
Face Recognition System ◽  
Cosine Distance

Due to, the great electronic development, which reinforced the need to define people's identities, different methods, and databases to identification people's identities have emerged. In this paper, we compare the results of two texture analysis methods: Local Binary Pattern (LBP) and Local Ternary Pattern (LTP). The comparison based on comparing the extracting facial texture features of 40 and 401 subjects taken from ORL and UFI databases respectively. As well, the comparison has taken in the account using three distance measurements such as; Manhattan Distance (MD), Euclidean Distance (ED), and Cosine Distance (CD). Where the maximum accuracy of the LBP method (99.23%) is obtained with a Manhattan and ORL database, while the LTP method attained (98.76%) using the same distance and database. While, the facial database of UFI shows low quality, which is satisfied 75.98% and 73.82% recognition rates using LBP and LTP respectively with Manhattan distance.

scholarly journals Discrete Wavelet Transform with Eigenface to Enhance Face Recognition Rate

2018 ◽  
Vol 7 (4) ◽  
pp. 9 ◽  
Author(s):  
Shakir F. Kak ◽  
Firas M. Mustafa ◽  
Pedro R. Valente
Keyword(s):  
Face Recognition ◽  
Recognition Rate ◽  
Training Image ◽  
Distance Measures ◽  
Manhattan Distance ◽  
Face Identification ◽  
Discrete Wavelet ◽  
Training Images ◽  
Matching Process ◽  
Cosine Distance

In a recent past, face recognition was one of the most popular methods and successful application of image processing field which is widely used in security and biometric applications. The innovation of new approaches to face identification technologies is continuously subject to building much strong face recognition algorithms. Face recognition in real-time applications has been fast-growing challenging and interesting. The human face identification process is not trivial task especially different face lighting and poses are captured to be matched. In this study, the proposed method is tested using a benchmark ORL database that contains 400 images of 40 persons as the variant posse, lighting, etc. Discrete avelet Transform technique is applied on the ORL database to enhance the accuracy and the recognition rate. The best recognition rate result obtained is 99.25%, when tested using 9 training images and 1 testing image with cosine distance measurement. The recognition rate Increased when applying 2-level of DWT with the bior5.5 filter on training image database and the test image. For feature extraction and dimension reduction, PCA is used. Euclidean distance, Manhattan distance, and Cosine distance are Distance measures used for the matching process.

Similarity Functions for Face Recognition

2015 ◽  
pp. 73-79
Keyword(s):  
Fuzzy Logic ◽  
Face Recognition ◽  
Similarity Functions ◽  
Cosine Distance

Similarity functions are not distances, but functions aimed to evaluate the similarity between two objects. Some of them relate to some other previously explained measures, such as cosine distance. Others are statistical or probabilistic, or rely on fuzzy logic. It has not been possible to provide a comprehensive table with recognition rates, as the data were to different to be compared.

scholarly journals Residual Compensation Networks for Heterogeneous Face Recognition

2019 ◽  
Vol 33 ◽  
pp. 8239-8246 ◽  
Author(s):  
Zhongying Deng ◽  
Xiaojiang Peng ◽  
Yu Qiao
Keyword(s):  
Neural Networks ◽  
Face Recognition ◽  
Learning Strategies ◽  
Convolutional Neural Networks ◽  
Network Architecture ◽  
State Of The Art ◽  
The Other ◽  
Training Images ◽  
Heterogeneous Face Recognition ◽  
Cosine Distance

Heterogeneous Face Recognition (HFR) is a challenging task due to large modality discrepancy as well as insufficient training images in certain modalities. In this paper, we propose a new two-branch network architecture, termed as Residual Compensation Networks (RCN), to learn separated features for different modalities in HFR. The RCN incorporates a residual compensation (RC) module and a modality discrepancy loss (MD loss) into traditional convolutional neural networks. The RC module reduces modal discrepancy by adding compensation to one of the modalities so that its representation can be close to the other modality. The MD loss alleviates modal discrepancy by minimizing the cosine distance between different modalities. In addition, we explore different architectures and positions for the RC module, and evaluate different transfer learning strategies for HFR. Extensive experiments on IIIT-D Viewed Sketch, Forensic Sketch, CASIA NIR-VIS 2.0 and CUHK NIR-VIS show that our RCN outperforms other state-of-the-art methods significantly.

Cosine-Based Distances, Correlations, and Angles for Face Recognition

2015 ◽  
pp. 47-55
Keyword(s):  
Face Recognition ◽  
Three Dimensional ◽  
Correlation Coefficients ◽  
Point Clouds ◽  
Jaccard Index ◽  
Gabor Wavelets ◽  
Feature Vectors ◽  
Cosine Distance

The cosine distance compares the feature vectors of two images by returning the cosine of the angle between two vectors. Other cosine- and angle-based measures are here presented, including Tanimoto dissimilarity and Jaccard index, together with other correlations; they have been employed in algorithms relying on PCA, ICA, NN, and Gabor wavelets, especially on bi-dimensional facial data. Only correlation coefficients have been applied on three-dimensional point clouds.

Face Inversion Effect Emerges Under Critical Configural Discrepancy

2010 ◽  
Vol 69 (3) ◽  
pp. 161-167 ◽  
Author(s):  
Jisien Yang ◽  
Adrian Schwaninger
Keyword(s):  
Face Recognition ◽  
Face Processing ◽  
Underlying Mechanism ◽  
Inversion Effect ◽  
Configural Processing ◽  
Major Contributor ◽  
Face Inversion Effect ◽  
Configural Information ◽  
Face Inversion ◽  
The Face

Configural processing has been considered the major contributor to the face inversion effect (FIE) in face recognition. However, most researchers have only obtained the FIE with one specific ratio of configural alteration. It remains unclear whether the ratio of configural alteration itself can mediate the occurrence of the FIE. We aimed to clarify this issue by manipulating the configural information parametrically using six different ratios, ranging from 4% to 24%. Participants were asked to judge whether a pair of faces were entirely identical or different. The paired faces that were to be compared were presented either simultaneously (Experiment 1) or sequentially (Experiment 2). Both experiments revealed that the FIE was observed only when the ratio of configural alteration was in the intermediate range. These results indicate that even though the FIE has been frequently adopted as an index to examine the underlying mechanism of face processing, the emergence of the FIE is not robust with any configural alteration but dependent on the ratio of configural alteration.

Perceptual Effects and Recollective Experience in Face Recognition

2005 ◽  
Vol 52 (3) ◽  
pp. 224-231 ◽  
Author(s):  
Chrisanthi Nega
Keyword(s):  
Face Recognition ◽  
Recognition Memory ◽  
Congruency Effect ◽  
Facial Recognition ◽  
Memory Systems ◽  
Study Time ◽  
Perceptual Fluency ◽  
Recollective Experience ◽  
Trace Strength ◽  
Size Congruency

Abstract. Four experiments were conducted investigating the effect of size congruency on facial recognition memory, measured by remember, know and guess responses. Different study times were employed, that is extremely short (300 and 700 ms), short (1,000 ms), and long times (5,000 ms). With the short study time (1,000 ms) size congruency occurred in knowing. With the long study time the effect of size congruency occurred in remembering. These results support the distinctiveness/fluency account of remembering and knowing as well as the memory systems account, since the size congruency effect that occurred in knowing under conditions that facilitated perceptual fluency also occurred independently in remembering under conditions that facilitated elaborative encoding. They do not support the idea that remember and know responses reflect differences in trace strength.

Sign in / Sign up

Export Citation Format

Share Document