scholarly journals Lightweight and Resource-Constrained Learning Network for Face Recognition with Performance Optimization

Sensors ◽  
2020 ◽  
Vol 20 (21) ◽  
pp. 6114
Author(s):  
Hsiao-Chi Li ◽  
Zong-Yue Deng ◽  
Hsin-Han Chiang
Keyword(s):  
Face Recognition ◽  
Performance Optimization ◽  
Learning Network ◽  
Key Points ◽  
Reference Images ◽  
In The Wild ◽  
Constrained Learning ◽  
The Difference ◽  
Computational Resources ◽  
High Degree

Despite considerable progress in face recognition technology in recent years, deep learning (DL) and convolutional neural networks (CNN) have revealed commendable recognition effects with the advent of artificial intelligence and big data. FaceNet was presented in 2015 and is able to significantly improve the accuracy of face recognition, while also being powerfully built to counteract several common issues, such as occlusion, blur, illumination change, and different angles of head pose. However, not all hardware can sustain the heavy computing load in the execution of the FaceNet model. In applications in the security industry, lightweight and efficient face recognition are two key points for facilitating the deployment of DL and CNN models directly in field devices, due to their limited edge computing capability and low equipment cost. To this end, this paper provides a lightweight learning network improved from FaceNet, which is called FN13, to break through the hardware limitation of constrained computational resources. The proposed FN13 takes the advantage of center loss to reduce the variations of the between-class features and enlarge the difference of the within-class features, instead of the triplet loss by using FaceNet. The resulting model reduces the number of parameters and maintains a high degree of accuracy, only requiring few grayscale reference images per subject. The validity of FN13 is demonstrated by conducting experiments on the Labeled Faces in the Wild (LFW) dataset, as well as an analytical discussion regarding specific disguise problems.

scholarly journals FT-GAN: Face Transformation with Key Points Alignment for Pose-Invariant Face Recognition

Electronics ◽  
2019 ◽  
Vol 8 (7) ◽  
pp. 807 ◽  
Author(s):  
Weiwei Zhuang ◽  
Liang Chen ◽  
Chaoqun Hong ◽  
Yuxin Liang ◽  
Keshou Wu
Keyword(s):  
Face Recognition ◽  
Generative Adversarial Networks ◽  
Facial Features ◽  
Adversarial Networks ◽  
Key Points ◽  
In The Wild ◽  
Novel Method ◽  
Face Synthesis ◽  
Unconstrained Face

Face recognition has been comprehensively studied. However, face recognition in the wild still suffers from unconstrained face directions. Frontal face synthesis is a popular solution, but some facial features are missed after synthesis. This paper presents a novel method for pose-invariant face recognition. It is based on face transformation with key points alignment based on generative adversarial networks (FT-GAN). In this method, we introduce CycleGAN for pixel transformation to achieve coarse face transformation results, and these results are refined by key point alignment. In this way, frontal face synthesis is modeled as a two-task process. The results of comprehensive experiments show the effectiveness of FT-GAN.

scholarly journals Deep Convolutional Neural Network-Based Approaches for Face Recognition

2019 ◽  
Vol 9 (20) ◽  
pp. 4397 ◽  
Author(s):  
Soad Almabdy ◽  
Lamiaa Elrefaei
Keyword(s):  
Neural Network ◽  
Face Recognition ◽  
Convolutional Neural Network ◽  
Performance Optimization ◽  
Large Scale ◽  
Visual Recognition ◽  
Recognition Accuracy ◽  
Support Vector ◽  
Svm Classifier ◽  
In The Wild

Face recognition (FR) is defined as the process through which people are identified using facial images. This technology is applied broadly in biometrics, security information, accessing controlled areas, keeping of the law by different enforcement bodies, smart cards, and surveillance technology. The facial recognition system is built using two steps. The first step is a process through which the facial features are picked up or extracted, and the second step is pattern classification. Deep learning, specifically the convolutional neural network (CNN), has recently made commendable progress in FR technology. This paper investigates the performance of the pre-trained CNN with multi-class support vector machine (SVM) classifier and the performance of transfer learning using the AlexNet model to perform classification. The study considers CNN architecture, which has so far recorded the best outcome in the ImageNet Large-Scale Visual Recognition Challenge (ILSVRC) in the past years, more specifically, AlexNet and ResNet-50. In order to determine performance optimization of the CNN algorithm, recognition accuracy was used as a determinant. Improved classification rates were seen in the comprehensive experiments that were completed on the various datasets of ORL, GTAV face, Georgia Tech face, labelled faces in the wild (LFW), frontalized labeled faces in the wild (F_LFW), YouTube face, and FEI faces. The result showed that our model achieved a higher accuracy compared to most of the state-of-the-art models. An accuracy range of 94% to 100% for models with all databases was obtained. Also, this was obtained with an improvement in recognition accuracy up to 39%.

scholarly journals PF -cpGAN: Profile to Frontal Coupled GAN for Face Recognition in the Wild

2020 ◽  
Author(s):  
Fariborz Taherkhani ◽  
Veeru Talreja ◽  
Jeremy Dawson ◽  
Matthew C. Valenti ◽  
Nasser M. Nasrabadi
Keyword(s):  
Face Recognition ◽  
In The Wild

scholarly journals Low-Resolution Face Recognition in the Wild via Selective Knowledge Distillation

2019 ◽  
Vol 28 (4) ◽  
pp. 2051-2062 ◽  
Author(s):  
Shiming Ge ◽  
Shengwei Zhao ◽  
Chenyu Li ◽  
Jia Li
Keyword(s):  
Face Recognition ◽  
Low Resolution ◽  
In The Wild ◽  
Knowledge Distillation

Gebruiksaspecten van het uitroeppartikel wat in het Nederlands en Fries

2020 ◽  
Vol 136 (4) ◽  
pp. 189-209
Author(s):  
Henk Wolf
Keyword(s):  
The Core ◽  
The Difference ◽  
Lexical Phrase ◽  
High Degree ◽  
Lexical Phrases

Abstract Both Dutch and (West) Frisian make use of the exclamative particle wat (‘how’), that adds an element of surprise about a high degree of something to the semantics of the sentence. In this paper I will first show the similarities between the use of the particle in the two languages. I will demonstrate that, in Dutch, its use is largely confined to constructions that are semantically scalable, whereas in Frisian this restriction is far less strict. I will explain the difference by showing that Dutch wat is a syntactic amplifier of lexical phrases, whereas Frisian wat has developed into a pragmatic amplifier of the core predicate. I will try to account for that difference by showing how homophonous words absent in Dutch are likely to have influenced the use of Frisian wat, and how Dutch prosody strengthens the connection between wat and the amplified lexical phrase, whereas Frisian prosody weakens it. Finally, I will show that the system described as ‘Frisian’ is occasionally found in varieties of Dutch too

scholarly journals Amplitude response of a telemetered seismic system from seismometers to digital acquisition system

1995 ◽  
Vol 38 (1) ◽  
Author(s):  
R. Di Giovambattista ◽  
S. Barba ◽  
A. Marchetti
Keyword(s):  
Digital Signals ◽  
Amplitude Response ◽  
Automatic Mode ◽  
Steady State Method ◽  
Sinusoidal Current ◽  
Fully Automatic ◽  
The Difference ◽  
High Degree ◽  
Seismic System

Automated amplitude response of the complete seismometer, telemetry and recording system js obtaiued trom sinusoidal inputs to the calibration coil. Custom-built software was designed to perform fully automatic cali- bration analyses of the digital signals. In this paper we describe the signals used for calibration and interactive and batch procedures designed to obtain calibration functions in automatic mode. By using a steady-state method we reach a high degree of accuracy in the determination of both the frequency and amplitude of the \ignal. The only parameters required by this procedure are the seismometer mass, the calibration-coil constant and the intensity of the current injected into the calibration coil. This procedure is applicable to telemetered seismic systems and represents an optimization of the processing time. The software was designed to requjre no modification" jf the device used to generate the sinusoidal current should change. In particular, it is possi- ble to changc the number of monotrequcncy packages transmitted to the calibration coil with the on]y restric- tion that the difference between the frequency of two consecutjve packages be greater than 5%; for these rea- sons the procedure is expected to be usefu] for the seismological community. The paper inc]udes a generaI de- scription of thc designing criteria, and of the hardware and software architecture, as well as an account of thc system's performancc during a two year period of operation.

scholarly journals Skolans statliga uppdrag

2018 ◽  
Vol 23 (5) ◽  
pp. 43-60 ◽  
Author(s):  
Gunnar Berg
Keyword(s):  
Social Norms ◽  
Liberal Education ◽  
Cultural Reproduction ◽  
School Activities ◽  
Societal Values ◽  
Production Factors ◽  
Societal Needs ◽  
Swedish Society ◽  
The Difference ◽  
High Degree

What tasks constitute the basis of the mission that the school as an institution is expected to fulfill in the Swedish society? In this article, I attempt to problematize that question. Few would oppose the notion that the school’s mission includes passing on knowledge as well as social norms. But what kind of knowledge should be put first in the day-to-day activities, and which social norms can and should be given priority? I argue that the school’s mission can be divided into two basic categories, here labeled competence development and cultural reproduction. The aim of the former is to educate the students with present and future societal needs in mind, focusing on growth ­– i.e. preparing the future labor force, taking macroeconomic and production factors into account. The task of cultural reproduction, on the other hand, is based on the notion of liberal education, where the aim is to reproduce societal values such as cultural heritage, democracy and citizenship. The difference between these two categories and their respective goals creates a number of tensions, resulting in a high degree of complexity in the daily school activities. Nevertheless, there are areas where the two categories at least partly intersect, meaning that the possibility of reaching a consensus does exist.

Production of Melanin from Aspergillus fumigatus as an Avirulence Factor

2019 ◽  
Vol 10 (04) ◽  
pp. 651-654
Author(s):  
Anmar Sael Hussein ◽  
Saja Jamal Noman
Keyword(s):  
Ir Spectra ◽  
Aspergillus Fumigatus ◽  
Wave Numbers ◽  
Quantitative Examination ◽  
Wave Range ◽  
The Difference ◽  
Chemical Characters ◽  
Chemical Groups ◽  
High Degree ◽  
Degree Of Similarity

In this quantitative examination of melanin was detected in 41 isolates of fumigatus. It was found that the significance of the difference in which the production of melanin (p ˂ 0.001) was the range of melanin extracted from fumigatus between (1.2 - 3.1 mg/mL). The isolates AFU1, AFU14, AFU29, AFU30, and AFU41 indicated that the maximum production ranged between (2.9-3.1 mg/mL). 2 hours at 100°C. It has similarities with standard melanin characters, and the same chemical characters of melanin extracted from the hymen of Bjerkandara adusta have been achieved according to a study (14) meaning there is similarity with standard melanin. The overlay of IR spectra of synthetic melanin extracted from fungal cultures showed a high degree of similarity. Purification of melanin from A. fumigatus. Wave numbers shown in 3381, 2927, 1867, 1531, 1404, 1073 and 651 cm-1 The wave range can be traced to the following chemical groups: 3381 cm-1 attributed to OH bonds, 2927 cm-1 to HC or HC = 0 bonds 1404 (C-CH3), and 651 cm -1 to (acyclic) CH2 bonds.

Sign in / Sign up

Export Citation Format

Share Document