Paravision

Paravision is a developer of a facial recognition platform, powered by computer vision and used by security device manufacturers, solution providers, systems integrators, and financial services firms to ensure identity. The company's facial recognition systems have ranked in the top ten of over fifty accuracy benchmarks across the 1:1 and 1:N NIST Face Recognition Vendor Tests. The platform offered by the company uses artificial intelligence at the edge and cloud-ready solutions to offer modular, scalable computer-vision solutions.

Paravision offers a computer vision developer toolset for a range of face recognition applications. The company offers face recognition SDKs and APIs to offer facial recognition software that is deployable in the cloud, on-premises, or at the edge. The toolset is capable of the following features:

• Face detection
• Face verification 1:1
• Face identification 1:N
• Profile face matching
• Real-time streaming video
• Liveness and anti-spoofing

The SDK is available for Python and C++ integrations, while the API is available for on-premise integration in data centers or in the cloud through Docker Containers. The solution has operating system support for Windows and Linux, and the company says it is optimized for Nvidia, Intel, and OpenVINO.

Built on Paravision's face recognition and computer vision toolset, the company offers a version of their facial recognition toolset designed for identifying and ensuring COVID-19 restrictions and regulations. This can include verifying identity and ensuring personal safety and enterprise security for workplaces, airports, and in retail. The toolset's capabilities include

• mask detection,
• face recognition with masks,
• integration with thermal detection, and
• person detection with tracking.

Paravision offers a suite of activity recognition tools with automated recognition of advanced activities and interactions. This toolset is intended to work with standard IP camera streams, offers performance in variable environments, is capable of using standard CPU and GPU compute systems, and offers a modular and open architecture. The activity recognition tools can be used for safety, security, and identity, with capabilities for

• object detection,
• attribute analysis,
• object tracking, and
• activity recognition.

Paravision offers its facial recognition toolset for access control, including biometric access using facial recognition to offer touchless access without need for keys, fobs, or phones, for a non-intrusive and intuitive access control for offices, campuses, events, or facilities. This can integrate video security, logical access, and HR systems for integrated identity across an enterprise. The face recognition is tested for angle and light variability; visual obstructions and occlusions; age, race, and gender diversity; low-quality video image; and profile views. Access control solutions are capable of the following:

• Face detection
• 1:1 face verification
• 1:N face identification
• Spoof detection
• Edge processing
• Real-time streaming

Paravision offers video security applications for their facial recognition toolset to help create safe and trusted environments using facial recognition and artificial intelligence-based computer vision tools. The toolset offers the capability for connecting a holistic view of identities, activities, and events in relation to the rest of a business. The video security toolset is capable of

• face detection,
• edge processing,
• 1:N face identification, and
• real-time streaming.

Paravision offers its facial recognition and computer vision systems for payments processing in order to provide touchless, efficient and personalized experiences. The company suggests the tool can offer secure identity verification to increase the speed of checkout, authorize banking transactions, and provide personalized offers to customers. This can be done through linking Paravision's tools with mobile, in-store, and curbside touch points. The company's system has been tested for a broad user demographics and environmental conditions to provide consistency and security for users and partners. The company's payments system is capable of the following:

• 1:1 face verification
• 1:N face identification
• Spoof detection
• Edge processing

Paravision offers identity and biometric verification programs for government and enterprise identity programs. The identity program is developed to be deployable on-premise or in the cloud, with the intention to enable a flexible solution for identity programs. This can offer a biometric system for large-scale identity programs, including voter registration, border facilitation, and social services. The system can be used for

• face detection,
• 1:1 face verification,
• 1:N face identification, and
• spoof detection.

Originally founded in 2013 as Ever, the company offered Everalbum, a cloud photo storage application. However, in 2020 the company announced it was shutting down its photo storage services, citing competition from the default services offered by Apple and Google as the cause. During that time, the company built Ever AI, a facial recognition application, which was offered through the Ever API to business clients, which have included law enforcement and the military.

In 2019, Ever AI formally rebranded as Paravision after the company's facial recognition achieved the top spot on the 1:1 Leaderboard against eighty-seven other vendors in the NIST Face Recognition Vendor Test. Previous to renaming, the company's facial recognition achieved top spot in the FRVT Wild Images category, and was one of the companies to score a top ten result in each category.

In 2020, Paravision came under scrutiny by the US Federal Trade Commission (FTC) based on complaints from users and The American Civil Liberties Union (ACLU) that Paravision had used photos from users to develop the facial recognition technology.

The FTC claimed Paravision made use of the pictures uploaded to its Ever service without permission, after the company discovered that Ever as a free cloud storage service for consumers would not make as much money as expected. Further, the FTC claims the service was active by default, except in Illinois, Texas, Washington and the European union, where laws about personal data and facial recognition require users to opt in to those services, rather than be opted in by default. The FTC also claimed that Paravision did not delete images from closed accounts, which the company said it would do in its terms.

In January 2021, FTC announced they had reached a settlement with Paravision. Under the terms of the agreement, the company was required to delete photos and videos of users with deactivated accounts as well as any facial recognition algorithms developed with users' photos or videos. This includes deleting all "face embedding," described as "data reflecting facial features used for facial recognition purposes" derived from users' photos without express user consent.

The company has previously stated it did not share personal user data, and the company's privacy policy said users' photos could be used to improve and train the company's facial recognition technologies. In the settlement, the FTC did not fine the company or offer any other fiscal punishment. Since the ruling, Paravision has appointed a Chief AI Ethics Advisor and published a set of AI Principles for the ethical development and use of face recognition and related technologies.