Enrollment is the process of detecting faces in an image (or video frame) and creating "templates" that encode the identifying characteristics of each face. The faster the enrollment speed, the less computing power required to support a face recognition application. If your face recognition application requires even faster enrollment speeds, please inquire about our ROC Embedded algorithm. Amongst the median submitted to NIST FRVT Ongoing, the ROC SDK is typically 2x faster.
The comparison speed is how long it takes to measure the similarity between two templates. The faster the comparison speed, the faster the verification or search results. Amongst the top 100 algorithms submitted to NIST FRVT Ongoing, the ROC SDK is typically 13x faster.
Speed is measured on Intel Xeon CPU E5-2630 v4 @ 2.20GHz using a single thread. Source: NIST FRVT Ongoing.
The template size is the amount of storage space required to save a template extracted in the enrollment process. The ROC SDK's small template size enables us to search large databases on a smartphone FASTER than some of our competitors can on a server.
Memory footprint is the total hard disk space that is required to store the ROC SDK software libraries. At 155MB, the standard ROC SDK memory footprint is one of the smallest on the market. At 7MB, the ROC Embedded memory footprint opens the door for mobile applications with powerful facial recognition capabilities embedded into the app and running on the edge device, without a data connection.
The ROC SDK is a fully native solution, with minimal system requirements and no internet is connection required. The ROC SDK supports all major operating systems and computer architectures. The ROC SDK exposes a C API with wrappers for Python, Java, C# and Go. A robust command line interface even allows constructing systems from shell scripts.
Perform 1:1 verification with industry leading accuracy.
Search millions of faces on a mobile device, or billions on a server, in seconds.
Locate faces in photos and video with industry leading detection speed.
Track eye and chin locations.
Group unlabeled faces by identity.
Predict a face likelihood to generate a successful match.
Analyze a face to predict its age, ethnicity and gender.
Analyze a face for anger, disgust, fear, joy, neutral, sadness or surprise expressions.
Process videos in real-time using a single core on a mobile device or server.
Determine where a person is looking relative to the camera using roll, yaw and pitch measurements.
Detect print and digital screen replay attacks using a single image.
Count persons in a video stream seamlessly.
Measure template similarity within the encrypted domain.
Detect eyeglasses, sunglasses, or no glasses.
Optionally enroll images on a GPU.
Determine if a face is of a cartoon, a painting or a human.
There is a misperception that face recognition algorithms do not work on persons of color, or are otherwise inaccurate in general. This is not true. The truth is that across a wide range of applications, modern face recognition algorithms achieve remarkably high accuracy on all races, and accuracy continues to improve at an exponential rate.
When automated face recognition technology is used for analyzing streaming video, an important question is: how much computer hardware is needed? The hardware required to process video depends on several factors which will be discussed in this article.
In this article we explain how to factor in the computational demand for template comparison in video processing applications. While this task is not as computational burdensome as template generation, for larger-scale applications it can become meaningful.
Law enforcement primarily uses face recognition as a post-incident forensic tool to enable detectives and analysts to generate investigative leads in violent and harmful crimes. In this article we explain how forensic face recognition works, and how it is used by law enforcement in this country.
The use of automated face recognition in law enforcement is one of the most powerful tools available in today’s law enforcement investigations, and delivers substantial benefits to society without any documented cases of law enforcement misuse.
This article will equip you with the knowledge to assess the efficiency requirements of your face recognition system. In turn, you will be able to factor this important consideration into your procurement process and potentially eliminate certain algorithms before the time consuming step of performing internal evaluations.
Perhaps no technology is improving as rapidly as automated face recognition. For example, over the last two years Rank One has reduced the False Non-Match Rate of our algorithm by over 10x. Other face recognition vendors are similarly improving their accuracy at a rapid pace. This article highlights the importance of an evergreen licensing approach to ensure you remain at the cutting edge of facial recognition capabilities.
Face recognition technology is rapidly expanding as a convenience technology that allows quick and secure access to sensitive systems and facilities, and for people to ubiquitously interact with their environments. This article highlights the most prominent commercial applications of face recognition technology that are emerging.
Choosing a face recognition algorithm that meets your accuracy requirements can be a daunting process. We simplify this process with a straightforward guide on how to measure algorithm accuracy and determine which algorithms are viable for your application.
Automated face recognition algorithms rely on highly complex mathematical models, but at a high level many of the techniques performed are rather easy to understand. In this article we provide a high level guide of how automated face recognition algorithms work.