Domain Adaptation for Unconstrained Ear Recognition with Convolutional Neural Networks DOI Creative Commons
Solange Ramos-Cooper, Guillermo Cámara-Chávez

CLEI electronic journal, Journal Year: 2022, Volume and Issue: 25(2)

Published: May 24, 2022

Ear recognition has gained attention within the biometrics community recently. images can be captured from a distance without contact, and explicit cooperation of subject is not needed. In addition, ears do suffer extreme change over time are affected by facial expressions. All these characteristics convenient when implementing surveillance security applications. At same time, applying any Deep Learning (DL) algorithm usually demands large amounts samples to train networks. Thus, we introduce large-scale database explore fine-tuning pre-trained Convolutional Neural Networks (CNN) adapt ear domain taken under uncontrolled conditions. We built an dataset VGGFace profiting face field. Moreover, according our experiments, adapting model leads better performance than using trained on general image recognition. The efficiency models been tested UERC achieving significant improvement around 9\% compared approaches in literature. Additionally, score-level fusion technique was explored combining matching scores two models. This resulted 4\% more. Open-set close-set experiments have performed evaluated Rank-1 Rank-5 rate metrics

Language: Английский

Machine learning in identity and access management systems: Survey and deep dive DOI

Sara Aboukadri,

Aafaf Ouaddah, Abdellatif Mezrioui

et al.

Computers & Security, Journal Year: 2024, Volume and Issue: 139, P. 103729 - 103729

Published: Jan. 23, 2024

Language: Английский

Citations

26
Deep Neural Network Based Vehicle Detection and Classification of Aerial Images DOI Creative Commons
Sandeep Kumar, Arpit Jain,

Shilpa Rani

et al.

Intelligent Automation & Soft Computing, Journal Year: 2022, Volume and Issue: 34(1), P. 119 - 131

Published: Jan. 1, 2022

The detection of the objects in ariel image has a significant impact on field parking space management, traffic management activities and surveillance systems. Traditional vehicle algorithms have some limitations as these are not working with complex background small size object bigger scenes. It is observed that researchers facing numerous problems classification, i.e., complicated background, vehicle’s modest size, other similar visual appearances correctly addressed. A robust algorithm for classification been proposed to overcome limitation existing techniques this research work. We propose an based Convolutional Neural Network (CNN) detect classify it into light heavy vehicles. performance approach was evaluated using variety benchmark datasets, including VEDAI, VIVID, UC Merced Land Use, Self database. To validate results, various parameters such accuracy, precision, recall, error, F1-Score were calculated. results suggest technique higher rate, which approximately 92.06% VEDAI dataset, 95.73% VIVID 90.17% 96.16% dataset.

Language: Английский

Citations

58
Ensemble Deep Learning and Internet of Things‐Based Automated COVID‐19 Diagnosis Framework DOI Creative Commons
Anita S. Kini, A. Nanda Gopal Reddy, Manjit Kaur

et al.

Contrast Media & Molecular Imaging, Journal Year: 2022, Volume and Issue: 2022(1)

Published: Jan. 1, 2022

Coronavirus disease (COVID-19) is a viral infection caused by SARS-CoV-2. The modalities such as computed tomography (CT) have been successfully utilized for the early stage diagnosis of COVID-19 infected patients. Recently, many researchers deep learning models automated screening suspected cases. An ensemble and Internet Things (IoT) based framework proposed Three well-known pretrained are ensembled. medical IoT devices to collect CT scans, diagnoses performed on servers. compared with thirteen competitive over four-class dataset. Experimental results reveal that ensembled model yielded 98.98% accuracy. Moreover, outperforms all in terms other performance metrics achieving 98.56% precision, 98.58% recall, 98.75% F-score, 98.57% AUC. Therefore, can improve acceleration diagnosis.

Language: Английский

Citations

40
Unlocking the black box: an in-depth review on interpretability, explainability, and reliability in deep learning DOI
Emrullah Şahin, Naciye Nur Arslan, Durmuş Özdemir

et al.

Neural Computing and Applications, Journal Year: 2024, Volume and Issue: unknown

Published: Nov. 18, 2024

Language: Английский

Citations

10
A Comprehensive survey on ear recognition: Databases, approaches, comparative analysis, and open challenges DOI Creative Commons
Amir Benzaoui, Yacine Khaldi,

Rafik Bouaouina

et al.

Neurocomputing, Journal Year: 2023, Volume and Issue: 537, P. 236 - 270

Published: March 30, 2023

Language: Английский

Citations

18
A Survey of 3D Ear Recognition Techniques DOI
Iyyakutti Iyappan Ganapathi, Syed Sadaf Ali, Surya Prakash

et al.

ACM Computing Surveys, Journal Year: 2022, Volume and Issue: 55(10), P. 1 - 36

Published: Sept. 2, 2022

Human recognition with biometrics is a rapidly emerging area of computer vision. Compared to other well-known biometric features such as the face, fingerprint, iris, and palmprint, ear has recently received considerable research attention. The system accepts 2D or 3D images input. Since pose, illumination, scale all affect images, it evident that they impact performance; therefore, are employed address these issues. geometric shapes ears utilized rich improve accuracy. We present recent advances in several areas relevant provide directions for future research. To best our knowledge, no comprehensive review been conducted on using human recognition. This focuses three primary categories techniques: (1) registration-based recognition, (2) local global feature-based (3) combination (2). Based above categorization publicly available datasets, this article reviews existing techniques.

Language: Английский

Citations

21
An efficient ear recognition technique based on deep ensemble learning approach DOI
Ravishankar Mehta, Koushlendra Kumar Singh

Evolving Systems, Journal Year: 2023, Volume and Issue: 15(3), P. 771 - 787

Published: May 4, 2023

Language: Английский

Citations

13
TR-ICANet: A Fast Unsupervised Deep-Learning-Based Scheme for Unconstrained Ear Recognition DOI
Aicha Korichi,

Sihem Slatnia,

Oussama Aiadi

et al.

Arabian Journal for Science and Engineering, Journal Year: 2022, Volume and Issue: 47(8), P. 9887 - 9898

Published: Jan. 9, 2022

Language: Английский

Citations

18
A Systematic Literature Review on Human Ear Biometrics: Approaches, Algorithms, and Trend in the Last Decade DOI Creative Commons

Oyediran George Oyebiyi,

Adebayo Abayomi‐Alli, Oluwasefunmi Arogundade

et al.

Information, Journal Year: 2023, Volume and Issue: 14(3), P. 192 - 192

Published: March 17, 2023

Biometric technology is fast gaining pace as a veritable developmental tool. So far, biometric procedures have been predominantly used to ensure identity and ear recognition techniques continue provide very robust research prospects. This paper proposes identify review present for biometrics using certain parameters: machine learning methods, directions future research. Ten databases were accessed, including ACM, Wiley, IEEE, Springer, Emerald, Elsevier, Sage, MIT, Taylor & Francis, Science Direct, 1121 publications retrieved. In order obtain relevant materials, some articles excused criteria such abstract eligibility, duplicity, uncertainty (indeterminate method). As result, 73 papers selected in-depth assessment significance. A quantitative analysis was carried out on the identified works search strategies: source, technique, datasets, status, architecture. Quantitative Analysis (QA) of feature extraction methods studies with geometric approach indicating highest value at 36%, followed by local method 27%. Several architectures, Convolutional Neural Network, restricted Boltzmann machine, auto-encoder, deep belief network, other unspecified showed 38%, 28%, 21%, 5%, 4%, respectively. Essentially, this survey also provides various status existing in classifying related studies. taxonomy current methodologies system presented along publicly available occlussion pose sensitive black image dataset 970 images. The study concludes need researchers consider improvements speed security algorithms.

Language: Английский

Citations

10
Synthesizing multilevel abstraction ear sketches for enhanced biometric recognition DOI Creative Commons
David Freire-Obregón, João C. Neves, Žiga Emeršič

et al.

Image and Vision Computing, Journal Year: 2025, Volume and Issue: unknown, P. 105424 - 105424

Published: Jan. 1, 2025

Language: Английский

Citations

0