Vintra claims success tackling bias in AI facial recognition
Working to end racial and ethnic bias in artificial intelligence (AI)-based biometric facial recognition, Vintra, a provider of AI-powered video analytics solutions, released the results of a year-long effort to ensure that its AI platform “can equitably recognize and correctly identify faces across different races,” the San Jose, California-based company said in a statement.
“With facial recognition algorithms outperforming APIs from tech leaders and popular open-source FR algorithms,” Vintra said it is focusing “on eliminating racial bias from machine learning-powered analytics.”
Vintra’s AI machine learning-powered video analytics software is designed to reduce bias, which has “resulted in cutting the bias gap by more than two-thirds and surpassing the accuracy rates across most racial and ethnic identities of the leading commercially available face recognition algorithms from Microsoft and Amazon, and those of the leading open-source face recognition algorithms like ArcFace,” the firm stated.
“We built our data models from the ground up and didn’t violate privacy policies by building out our training data. We focused on quality data from day one and adhering to our core principles of integrity and trust. Simply put, it matters how we got here, and it shows in the results,” said Vintra CEO Brent Boekestein.
The company said its “analysis … was computed on the well-accepted ‘Racial Faces in the Wild’ (RFW) dataset and demonstrates how Vintra has the lowest bias results,” noting that the authors of the Cornell University paper, Racial Faces in-the-Wild: Reducing Racial Bias by Information Maximization Adaptation Network computed results on various ethnicities from Amazon and Microsoft’s face recognition API,” and “the average of those results is provided … as [a] reference.”
“Racial bias is an important issue in biometric, but has not been thoroughly studied in deep face recognition,” said the authors of the Cornell University paper, Mei Wang, Weihong Deng, Jiani Hu, Xunqiang Tao, and Yaohai Huang.
Using the dedicated Racial Faces in-the-Wild dataset, the researchers wrote, “we firmly validated the racial bias of four commercial APIs and four state-of-the-art algorithms.” Their solution was accomplished, they said, by “using deep unsupervised domain adaptation,” and proposing “a deep information maximization adaptation network (IMAN) to alleviate this bias by using Caucasian as source domain and other races as target domains. This unsupervised method simultaneously aligns global distribution to decrease race gap at domain-level, and learns the discriminative target representations at the cluster level.”
The researchers said, “A novel mutual information loss is proposed to further enhance the discriminative ability of network output without label information,” noting that “extensive experiments on RFW, GBU, and IJB-A databases show that IMAN successfully learns features that generalize well across different races and across different databases.”
Vintra explained the problem like this: “The public is largely concerned with the face recognition working properly when it is deployed, and Vintra strongly supports that concern. AI developed from machine learning is often based on huge, publicly available data sets that many AI solutions providers use to build their product.”
“Unfortunately,” the company said, “most Western-based facial recognition algorithms misidentify faces that aren’t Caucasian or lighter-skinned for two key reasons.”
First, it said “core data sets are populated by a super majority of white faces,” and secondly, “algorithms have been built and tested on these data sets for years. While algorithmic tweaks have tried to address this in the past, a poor representation of any given ethnicity in a data set cannot be overcome by algorithmic changes alone.”
Vintra outlined its solution saying it “has built and curated [its] own data set, pulling from over 76 countries and tens of thousands of identities with dozens of reference images for each identity in order to better represent Caucasian, African, Asian, and Indian ethnicities. This work has resulted in a much fairer balance with each group representing roughly 25 percent of the total data population and giving AI-powered video analytics a truer picture of what our world really looks like.”
Ensuring that the accuracy of Vintra’s facial recognition results endured in the top ten percent of solutions globally when tested on leading datasets like RFW, the firm said its “team set out to reduce the bias gap,” which is described as “the percentage delta between correctly identifying white faces and all other non-white identities.”
Publicly available academic algorithms have an eight percent difference in bias between black and white faces. In comparison, commercially available algorithms have a nine percent difference, Vintra said its results determined, adding that “some companies, notably Microsoft and Amazon, have a 12 percentage point difference when looking at white and black faces.”
Vintra said, “With initial testing of the new dataset and algorithm,” it has been able to close “the racial bias gap to 4.7 when comparing Caucasian and African-descent faces,” and that its “average accuracy across all non-Caucasian categories beats popular APIs from Amazon, Microsoft, and the leading open-sourced algorithms on the market.”
Vintra said in its statement that it “is committed to ensuring that facial recognition technology can be developed to equitably recognize all types of faces,” and that “the more fair and accurate the results of these solutions are, the more they will be accepted by society as a force for good.”
FulcrumAI, Vintra’s video analytics platform, is capable of transforming video from any type of camera into actionable, tailored, and trusted intelligence, the company claimed, adding that “powerful yet flexible, FulcrumAI can be deployed on-premises to augment human resources by leveraging live video feeds that deliver timely preventative alerts and situational awareness, or in the cloud as a powerful post-event investigation solution.”
The company added that its FulcrumAI platform also “provides powerful video analytics for private security professionals and public safety officials that can be custom-tuned for any environment.”
Vintra raised $4.8 million in funding last year and is backed by venture capital funding led by Bonfire Ventures and Vertex Ventures.