Deep learning facial recognition technology – WV-ASF950
Panasonic FacePRO can identify faces that are difficult to recognise within common video surveillance technology. These situations include reading faces at an angle of up to 45 degrees to the left or right or 30 degrees up or down, with a 90% accuracy rate when detecting faces partially hidden by sunglasses or face masks. FacePRO also has high accuracy reads in matching faces taken from up to ten-year-old passport images.
Specific faces can be registered in advance to send an alarm when they are detected. The faces of repeat shoplifters and wanted criminals,etc. can be registered in the facial recognition security/surveillance system from data recorded in the past.
Information can also be shared between facial recognition systems by importing generic photo data in the JPEG format, etc. Alarms can notify the operator by displaying pop-ups on the screen, emitting warning sounds, or flashing the camera on the map, etc.
Face images can be used to perform searches. For example, face images of suspicious people detected on the sales floor can be searched to track information on a timeline, including what time they entered the store and which sales floor they passed. Whether suspicious people have shoplifted, etc. after entering the store can also be immediately searched for and checked by the facial recognition security/surveillance system “FacePRO”.
One of the major advantages of the Facial Recognition Solution is the level of control it gives you from a central point of management.
Thanks to the direct connection between camera and server, you can control and gather insight from up to 20 remote cameras from one server, at the same time. With face matching and visual analysis carried out by the system in real time, your operator can easily access any recorded images related to that face or customer, via the GUI in their monitoring software.
The facial recognition software runs silently in your system, collecting data on each face that it detects; this data is then stored within an easily accessible database.
A user will then be able to access this database, and will be given the option to select a particular face. With this functionality, the user can choose to set an alarm to be sounded upon future detection of this subject, or simply track a particular person’s movements in chronological order throughout all of the cameras in the system.
People Counting, Age and Gender Statistics
The software also features an analytics section. In this area, a user is given the ability to analyse statistics such as people counting, and also allows them to detect age and gender. The system will then display the relevant information in an accessible manner.
Innovative Application: Attendance Taking System
– The deep learning technology used in the new software was jointly developed with the National University of Singapore and improves facial recognition performance by up to 500% compared to conventional systems. (Rejection rate reduced to 20% when the wrong person acceptance rate is set to 0.01 with the IJB-A face image dataset.)
A unique algorithm that combines deep learning, a machine learning method, with a similarity calculation method that suppresses errors, enables recognition in situations that were difficult with conventional facial recognition technology, such as when the face is angled (up to 45 degrees to the left or right or 30 degrees up or down), partially hidden with sunglasses or a surgical mask*5, or changed by aging.