Researchers Propose a Deep Learning-Based Face Recognition Technology with an Accuracy of 99.95% for Facial Recognition Even for a Person Wearing a Niqab

Computer algorithms are used by face recognition systems to identify certain recognizable features on a person’s face. The information is then transformed into a mathematical representation and contrasted with information on other faces gathered in a face recognition database. Examples of these features are the space between the eyes or the contour of the chin.

Face-recognition technology is quickly developing and used in various fields, including marketing, education, criminal investigation, security, and biometrics. Now, in addition to being able to identify the individual, it can also determine their facial expression. The limits of facial recognition software when a person’s face is partially hidden, as can happen when wearing a veil or protective face mask, are the subject of research published in the International Journal of Biometrics.

Full-face biometric identification has been the subject of a substantial amount of research. However, employing faces that are only partially visible, like veiled people, is difficult. In this study, the deep convolutional neural network (CNN) is used to extract characteristics from photographs of veiled people’s faces.

The researchers claim that their deep-learning technique for facial recognition is 99.95% correct, even when a person is wearing a niqab, which mostly hides the face except for the eyes. Age estimation and gender recognition by the algorithms are both 99.9% correct. Examining the eyes can identify a veiled person or wearing a COVID mask as happy or frowning with an accuracy of 80.9%. A database of 150 images was used for the tests, including 109 female and 41 male participants ranging in age from 8 to 78. The researcher used a deep con to Each layer of the recognition process in the neural network has 4096 features.

The team notes that its proof of concept, DeepVeil, used face-on photos of veiled people shot up close together with an internal picture database. The following phase will include working with a more varied collection of images captured in various contexts, including pictures taken at multiple angles. However, as algorithms and software have advanced, it is no longer necessary to have a clear face-on image to confirm a person’s identification as it was in the early days of traditional facial recognition systems. Therefore, it is conceivable that DeepVeil will experience the same thing with the proper strategy and continued development.

This Article is written as a research summary article by Marktechpost Staff based on the research paper 'DeepVeil: deep learning for identification of face, gender, expression recognition under veiled conditions'. All Credit For This Research Goes To Researchers on This Project. Check out the paper and reference article.

Please Don't Forget To Join Our ML Subreddit

I am consulting intern at MarktechPost. I am majoring in Mechanical Engineering at IIT Kanpur. My interest lies in the field of machining and Robotics. Besides, I have a keen interest in AI, ML, DL, and related areas. I am a tech enthusiast and passionate about new technologies and their real-life uses.