Clearview granted patent for building biometric training dataset by USPTO
A patent has been granted to Clearview AI for its method of gathering and preparing training data for face biometrics algorithms.
The U.S. Patent and Trademark Office issued the patent for ‘Scalable Training Data Preparation Pipeline And Efficient Distributed Trainer For Deep Neural Networks In Facial Recognition,’ which the company says enables it to create facial recognition that is highly accurate and bias-free from publicly available information.
“This distinction further cements Clearview AI’s intellectual property protection and lead in the artificial intelligence and facial recognition market,” says Clearview CEO Hoan Ton-That. “Clearview AI’s mission is to reduce bias in technology, and as a person of mixed race this is highly important to me.”
Ton-That and Clearview Vice President of Research Terence Liu discussed how clusters are built within training datasets and the importance of patenting the technology with Biometric Update in a recent interview.
Clearview’s approach to training data differs from that of some other facial recognition developers, who use datasets made up of celebrities that do not evenly represent all demographics.
“Incumbent companies sometimes wait and see with new technologies to see the viability and adoption of them in the marketplace, then copy innovations later once they have been proven to be valuable,” Ton-That adds. “These patents help protect us against a potential future competitor who would like to copy our facial recognition search engine, or our method for creating a highly accurate, bias-free facial recognition algorithm from large scale public internet datasets.”
Clearview says the success of its approach is reflected in its results in the NIST Face Recognition Vendor Test, including top-two results in the Wild category and better than 99 percent accuracy across all demographics.
The company was also awarded a patent for applying facial recognition to gather information from the public internet earlier this year.