Thinking outside the box
Every conversation about face recognition begins with ‘How accurate is it?’ and this is just not a simple question to answer. There are multiple factors and multiple variables within each use case which will impact the answer to the question. Fortunately, CyberExtruder has approached the challenge of matching faces in a very different way than everyone else. The cornerstone of this is our patented 2D to 3D transformational process which mitigates a lot of the variables that can negatively impact matching scores. Then layer in our proprietary deep convolutional neural networks, and the result is a face matching system which is uniquely fast and accurate by design, flexible in use and scalable for tomorrow because what we can uncover about a face as it transforms from 2D to 3D allows us to solve problems no one has yet considered.
Face recognition performed on video presents levels of difficulty not seen in controlled environments because subjects may be absent for extended periods, with cluttered backgrounds yielding false face detections in other algorithms. Subjects are not only in motion, but the direction they are moving is not optimal for face matching purposes. Motion can cause blurring which means that to be effective, algorithms have to track faces through time. This task is further complicated by busy backgrounds, variations in lighting and the possibility of faces being hidden (occluded) by objects or other people.
Now add in the likelihood of multiple people being present in a scene each one requiring simultaneous detection, tracking, and matching. Facial detail is degraded due to the lossy compression that is present in almost all video transmission and storage.
Our answer to all these challenges is Aureus 3D and CyberExtruder’s continuing program of research and development in advanced face recognition (FR) technologies. We believe we have developed the world’s fastest and most accurate face matching engine. Our performance superiority is based on our proprietary 2D to 3D face modeling technology which mitigates the challenges of pose, lighting, and expression as well as our use of convolutional neural networks which allow us to train our algorithms to deliver ever increasing performance.
But it’s not just about being the best at matching faces. Our approach optimizes the competitive difference of 3D over 2D technology and delivers real-time face matching is full frame rate video. Our 3D technology compensates for the pose, lighting, facial expressions, and vague images – which are the biggest constraints to conventional FR. Our 3D technology works with any video camera, and we provide the most accurate face tracking technology anywhere. Best of all, our software is designed for easy partner integration.