International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 11 Issue: 05 | May 2024
www.irjet.net
p-ISSN: 2395-0072
A Comparative Study of Classical and Modern Face Detection and Recognition Methods: Accuracy, Challenges, Efficiency and Performance Analysis Omkar Dengi , Assistant Professor, Dept. of Computer Science, Dr. D.Y. Patil ACS College Pimpri Pune, Maharashtra, India ---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Face detection and recognition are essential
Face recognition system has two main tasks: Face detection and Face recognition
components of computer vision with wide-ranging applications in security, surveillance, biometrics, and humancomputer interaction. This paper provides a thorough comparative analysis of classical and modern methods for face detection and recognition. Traditional approaches, founded on predefined rules and handcrafted features, established the groundwork for contemporary algorithms. Techniques such as Viola-Jones, HOG, Eigenfaces, Fisherface, LBP, and ASM are adept at extracting relevant features for face detection. Meanwhile, machine learning-based methods like SVM, k-NN, and Decision Trees excel in classifying regions for recognition tasks. In contrast, modern methodologies harness the power of deep learning and sophisticated techniques for superior performance. Models like RetinaFace, BlazeFace, ArcFace, CenterFace, InsightFace, DeepFace, FaceNet, SphereFace, and DeepID leverage deep neural networks to achieve enhanced accuracy and efficiency. Additionally, this paper delves into emerging trends such as 3D face recognition and cloud-based solutions like Amazon Rekognition. Overall, modern based methods outperform classical techniques in terms of accuracy and efficiency, especially when dealing with large datasets and challenging conditions.
Face detection is the process of locating human faces within images or video frames. The goal is to identify the presence and location of faces, typically represented as rectangular bounding boxes around detected faces. Face recognition goes beyond face detection, identifying and verifying individuals based on their unique facial features. It involves comparing detected faces with a database of known faces to determine their identity. Face recognition system involves verification and identification process. Face verification means a 1:1 match that compares a face images against a template face images whose identity is being claimed. Face identification involves comparing a query face image against all image templates in a face database, which is a 1: N problem. The process of face recognition can be illustrated in Fig1. Fig1: Face Recognition Process
Key Words:
Face detection , Face recognition, RetinaFace, BlazeFace, ArcFace, CenterFace, InsightFace, DeepFace, FaceNet, SphereFace, Machine learning, Amazon Rekognition, Deep Learning
1. INTRODUCTION Face detection and recognition are integral components of computer vision systems designed to identify and analyze faces within images or video streams. One of the most interesting and successful applications of pattern recognition and image analysis is face recognition, which stands as a prominent biometric authentication technique. Face detection and recognition are fundamental tasks in computer vision with widespread applications in security, surveillance, human-computer interaction, and more. This paper presents a comprehensive overview of the evolution of face detection and recognition techniques, from classical methods to modern deep learning-based approaches.
© 2024, IRJET
|
Impact Factor value: 8.226
In any face recognition system, the first step is to detect a face in an image. The main goal of face detection is to determine if there are any faces present in the image. If a face is detected, the system returns the location of each face in the image. In the face recognition process, the input image, also known as a probe, is compared with the database, referred to as a gallery. A match report is generated, followed by classification to determine the subpopulation to which new observations belong.
|
ISO 9001:2008 Certified Journal
|
Page 704