Security-oriented Face Detection Technology Utilizing Deep Learning Techniques Along with the CASIA Datasets
Iqra Yamin1, Yang Gaoming2, Marcel BAKALA3, Muhammad Asad Yamin4, Usama Masood5
1Iqra Yamin, Department of Computer Science and Engineering, Anhui University of Science and Technology, Huainan (Anhui), China.
2Yang Gaoming, Department of Computer Science and Engineering, Anhui University of Science and Technology, Huainan (Anhui), China.
3Marcel BAKALA, Department of Computer Science and Engineering, Anhui University of Science and Technology, Huainan (Anhui), China.
4Muhammad Asad Yamin, Department of Computational Science and Engineering, University of Rostock Germany.
5Usama Masood, Department of Mechanical Engineering, Anhui University of Science and Technology, Huainan (Anhui), China.
Manuscript received on 08 November 2023 | Revised Manuscript received on 17 November 2023 | Manuscript Accepted on 15 January 2024 | Manuscript published on 30 January 2024. | PP: 1-11 | Volume-12 Issue-5, January 2024 | Retrieval Number: 100.1/ijrte.E79700112524 | DOI: 10.35940/ijrte.E7970.12050124
Open Access | Editorial and Publishing Policies | Cite | Zenodo | OJS | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Recently, face recognition technology has become increasingly important for safety purposes. Masks are now required in most countries and are increasingly used. Public health professionals advise people to cover their faces outdoors to reduce COVID-19 transmission by 65%. Detecting people without masks on their faces is crucial. This has become widely used as face recognition outperforms PINs, passwords, fingerprints, and other safety verification methods. Sunglasses, scarves, caps, and makeup have made facial identification harder in recent decades. Thus, such masks impact facial recognition performance. Face masks also render traditional facial recognition technology ineffective for face authorisation, security checks, school monitoring, and opening cellphones and laptops. Thus, we proposed Masked Facial Recognition (MFR) to recognise both veiled and exposed-face individuals, allowing them to avoid removing their masks to verify their identities. This deep learning model was trained using the Inception ResNet V1. CASIA is responsible for preparing pictures and using LFW to validate models. Dlib creates masked datasets utilizing vision algorithms. An accuracy of approximately 96% was achieved using our three trained models. Thus, covered and uncovered recognition of faces, as well as detection techniques, can be easily applied in security and safety verification. These systems can be utilised in various settings, including airports, train stations, and other public areas, to enhance security and deter crime. Overall, deep learning within face recognition technology has significant potential for improving safety and security in various settings.
Keywords: CASIA Dataset, Dlib, face recognition, Masked Facial Recognition.
Scope of the Article: Computer Science and Its Applications