Oosto Deep Learning Improves Facial Recognition Accuracy, Expands Uses
New facial recognition systems are scalable and provide more accurate results, and the end user selects system performance parameters to meet the needs of their own applications. “We have developed algorithms that can be optimized in a variety of compute environments, whether in a small chip inside a peripheral device or on a large server,” said Ido Amidi, vice president of products and business. business development of Oosto. “It all depends on the needs of the customers. We can do it all without sacrificing performance. There is no significant loss of performance on the different platforms. The system’s “containerized environment” operates efficiently within a customer’s IT architecture. Facial Recognition Implementation Oosto, formerly AnyVision, was founded in 2016 on the basis of deep learning research on facial recognition. The focus was on how to use facial recognition to identify people in real world scenarios. Products began to market in 2017, targeting verticals that require identifying ‘bad actors’ or very important people (VIPs). The implementation of the technology involves the operation of the software in an on-premises device or server. a network video recorder or access control system. As Internet of Things (IoT) concepts spread across the security market, edge-based deployments will become more common, providing actionable insights in real time and avoiding an influx of unstructured video data. Ethical recognition Oosto helps users deploy, configure and calibrate the system; then, it is managed by the client according to his needs. Privacy features are built into the system as part of the company’s commitment to “ethical” facial recognition. Technology, in general, has changed a lot since the first implementations a decade ago did not perform as promised. Extended capabilities are fueled by developments in deep learning. Privacy Features Facial recognition can help end users identify people of interest, but is specifically designed not to violate anyone’s privacy. There are no databases or watchlists involved; the user usually compiles their selection of “bad actors” against which facial recognition algorithms can be compared. While Oosto provides a valuable tool, the end user customer decides how to deploy this tool in their business, explains Amidi. Oosto’s “activity detection” deploys algorithms to analyze images from videos and / or 3D cameras. Activity detection Recently, Oosto has adapted the approach to provide access control; also, including “vividness detection” which deploys algorithms to analyze images from video and / or 3D cameras to ensure that a face presented is not a printed image or mask. Their latest software release is a unique platform to recognize and deliver information about how people behave in physical spaces. The new functionality includes the ability to detect unknown individuals entering restricted areas. Neural Network Algorithm Improved facial recognition works well even for those who wear masks. A new neural network algorithm improves accuracy when identifying people wearing masks, and the system can alert if a person is not wearing a mask (to aid compliance). Extensive video forensics features, designed to speed up investigations, include the ability to perform in-depth searches of captured video related to bodily attributes, clothing color, and more. Recognition algorithms How well can Oosto recognize a face among 20 different people on a casino floor? It all depends on variables, such as number of people, lighting and viewing angle, picture quality and even network issues. Recognition algorithms are “trained” using data collected from real use cases; training takes place in the laboratory environment. The system can adapt to various lighting conditions, such as in a glazed hall in the morning versus in the afternoon. When people enter an area, for example, they can look at their smartphones, which would mean that some of their faces would not be visible. Such variables have an impact on precision. System testing requirements apply to each specific customer and can be adjusted and changed as needed Customizable system Each customer must decide their settings for how best to use the system, with the fewest acceptable false alarms , in real time existing infrastructures. The requirements for testing a system apply to each specific customer, and the sensitivity of the system and other operational factors can be adjusted and fine-tuned as needed, including at any time after a system is installed. Respond to the specified application “No client is the same, so each client’s system needs are unique,” Amidi explains. For example, some situations may have a higher tolerance for false alarms, such as when an operator is available and can make the final decision as to whether a face matches. In other situations, such as when a child goes missing, false alarms are a bigger problem and a quick response is especially essential. “What we offer out-of-the-box customers applies to any scenario, and customers can change the system with the click of a button to respond to a specific application,” Amidi explains. “We allow our customers to understand the pros and cons so that they can react in real time. »Identity Management Oosto can help improve the customer experience by identifying important customers. Oosto sells through system integrators and has a partner program with over 150 integrators worldwide certified to manage and install systems. Oosto targets Fortune 500-sized companies in financial services, gaming and retail. The system helps to create a safe environment free from “bad actors”. Alternatively, Oosto can help improve the customer experience by identifying important customers who deserve special treatment. The system can also provide an alert if an unauthorized person enters a restricted area; that is, anyone who is not an employee, registered visitor or contractor. Understanding the Technology In many parts of the world, facial recognition is widely accepted and used for applications such as payment and access control. Anxiety about technology, especially in the United States, is based on a lack of understanding. The public needs to be “better educated” on the subject, and Amidi expects the technology to become more socially accepted over time. “It is a tool, and it must be supervised, specifies Amidi. “From a technological point of view, we trust precision. It just needs to be explained better.