Review on Human Face Detection Using Deep Learning
Year: 2024
Type: article
Abstract: In the real world, the detection of human face objects is challenging and seems to be a complex task. Emotions play a vital role for every individual to understand their own self and others and honour it. Emotional intelligence (EI) is one's ability to perceive others to manage better relationships. Facial Detection is the primary key in EI and to achieve this potential, artificial intelligence helps in EI's needs. In today's world, domains like mentoring, photography, security systems, social media platforms, augmented reality and so on, use face detection as a tool for assessment and surveillance. The significance of this review lies in its multidisciplinary focus, bridging the gap between computer science and psychiatry. As soon as one must identify facial expressions, the process gets complex. The demand to provide a promising model to recognize faces with respect to computational cost, processing speed and execution time even in low-power devices has become more prominent in real time applications. By focusing on emotion detection and social behaviour analysis, these systems can offer clinicians novel ways to assess and monitor psychiatric patients, thus contributing to personalized treatment strategies. In this work, various deep learning algorithms are tested to detect the faces of different resolutions. MTCNN, also known as Multi-Task Cascaded CNN, specializes in detecting and learning features from facial images that relate to facial expressions. This proposed system aims to accurately detect faces in real time by extracting the feature points and feeding them to the model to recognize the landmarks as an efficient way for effective and precise facial expression identification, with potential uses in a number of fields, such as robotics, human-computer interface, sales, and healthcare. Hence, through experimental analysis and performance visualization, this work identifies the best deep learning model achieving better accuracy with feasible performance, speed and execution time on large datasets when compared to other models.
Author: A. Srinidhi, S. Bhavanisankari
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W6949556842
FWCI:
Cites:
Cited by:
Related to: