piXlogic facial recognition software, piXserve, is designed to identify and verify a person's identity by using biometrics to piece together facial features from an image or video. It uses and extends state-of-the-art AI and computer vision technologies to extract actionable information from these unstructured data sources. piXserve automatically scans and indexes media files and live video sources, assigning keywords to objects it recognizes to make this content searchable. The award-winning face detection software provides a compelling solution for the intelligence community and state and local law enforcement.
Facial recognition is a process of scanning a database of images or video frames to identify and verify a person’s face using biometrics. Historically, facial detection work has been carried out manually by law enforcement officials who had to pinpoint the coordinates of facial features to identify a suspect. However, computers are much faster and more reliable in facial analysis than humans. Computerized facial recognition technology can be programmed to provide possible matches in natural or uncontrolled environments even when the image is distorted. Automated facial recognition software can also produce matches when face characteristics have morphed due to aging or injury.
In essence, facial recognition technology computes the distance between distinctive details about an individual’s face, such as the eyes, nose, ears, eyebrows, etc. Since the eyes do not change position much over time, it’s the fixed reference to calculate point distance between the other features. However, eye localization does not vary much between individuals, meaning that measurements must be very precise to achieve the best outcomes. Otherwise, a small error will have significant impact on the calculations and accuracy of the results.
Given the complexity involved in measuring point distance between facial features, computerized facial recognition evolved to measure faces using texture descriptors. Texture calculations cover areas and are more robust than point distance calculations, allowing for a larger margin of error. However, texture is a function of scale, and it is still important to localize eyes. Even so, users can still obtain reasonable results even when eye localization is less than perfect since the program is taking in a wider range of variables to cross-check and calculate the probability of a match.
More recently, with the advent of convolutional neural networks, attention has shifted to using trained models that embed millions of parameters. These software models are able to incorporate metrics at different scales and orientations, and because a larger range of variables are covered, localization errors, while still important, can be tolerated over a broader range. Face detection and face recognition model are trained and calibrated over millions of examples. The accuracy that can be achieved even for challenging face matching situations is quite high.
From birth, people have an innate ability to detect and recognize faces, and the better you know a person, the easier it is to recognize them. However, that ability varies from individual to individual, and most often matching scenarios do not involve those familiar to examiners. Facial recognition software is quite accurate and can even surpass human face recognition by a fair margin, but a recent study found the best results were achieved when both facial recognition software and human experts worked together. Just like humans can improve skills with practice, computerized face detection algorithms can improve their accuracy. The more examples imported to a software-based facial recognition tool, the higher the likelihood of a precise match.
The beneficial applications of facial recognition technology range from safety applications to cybersecurity, with law enforcement being the earliest adopter. Without the level of automation that computerized facial recognition technology brings to the table, law enforcement in today’s world would be significantly less effective and efficient. There are additional use cases related to convenience and cybersecurity, such as using your face to unlock a cell phone, unlock a door or punch a time card. Soon, instead of storing and remembering passwords for all manners of websites, stores, banks and facilities, we can use our live face as the key to unlocking those services. Facial recognition can also help power safety applications, such as making sure the correct prescription is taken by the right person at a hospital or identifying the parents of a lost child on the street. The technology has become advanced and continues to develop, leaving the potential for universal adoption in the near future as more pieces of our lives shift to digital.
Image Search and Indexing
For more information, please contact us at AWS@carahsoft.com.