Trueface.ai shares principles for responsible deployment of facial recognition technology
Nearly five years after creating facial recognition doorbell Chui, the company’s creators are touting their Trueface.ai facial recognition software as part of the way forward for the technology through responsible deployment. The software is meant for edge deployments in highly constrained environments, and the company says it can identify spoof and hack attempts, TechCrunch reports.
The company initially existed to develop and sell a smart home doorbell equipped with facial recognition, which was unveiled in 2014, but the technology was not advanced enough to prevent the spoofs and presentation attacks necessary for a home access control application.
When they upgraded the software, commercial access control providers were interested in using it in their own hardware, so the team closed its manufacturing operation and spun Trueface out of Chui in June, 2017, releasing an API in July. A seed capital round led by Scout Ventures, with participation from Harvard Business Angels and GSV Labs raised $1.5 million, and the company integrated its software with the IFTTT platform at the beginning of 2018.
Trueface.ai CEO Shaun Moore writes in a blog post that the company pioneered new methods of collecting training data, including partnerships with non-profit organizations in Africa and Southeast Asia, to reduce algorithmic bias and ensure accuracy for people of different ethnicities and skin colors.
In the post, Moore outlines three fundamental principles the company operates on. The “Humanity First” principle includes a commitment to bias reduction, and a statement in partner contracts that humans will make the ultimate decisions in law enforcement use cases. The “Data Security Focused” principle is a commitment to keeping deployed software and customer data on the servers of Trueface.ai’s clients. The third principle is a commitment to “Total Transparency.”
Moore acknowledges that responsible deployment means different things to different stakeholders, but also writes: “Soon, computer vision will be a fundamental layer on every camera. When implemented responsibly, people will demand this technology for its daily benefits and utility, not fear it.”