DURHAM, N.C. — Video surveillance cameras recorded thousands of students walking the Duke University campus one day in March 2014 as part of a research project intended to help accelerate “multi-target, multi-camera tracking” (MTMC) systems.
The video data collected from the project would later be used by China and others to test facial recognition software, according to an article from Duke’s independent student newspaper, The Chronicle.
The recordings and their likenesses were put into a data set that was “placed on a public website, and [was] downloaded by academics, security contractors and military researchers around the globe,” the article states.
The footage was used to test and improve facial recognition technology that was mostly used for either private or public security purposes. The data set has been linked to the Chinese government’s use of facial technology to monitor ethnic minorities there, The Chronicle reported.
The data set and the project’s Duke website were taken down in April, after Microsoft came under fire for its facial recognition database that had more than 10 million images of roughly 100,000 people, according to the article.
The company’s database was exposed in a Financial Times investigation, in which the data set from Duke was also mentioned as “one of the most popular pedestrian recognition training sets.”
The article details:
In September 2016, researchers from Duke and the University of Modena and Reggio Emilia in Modena, Italy published a paper written to help accelerate progress with developing multi-target, multi-camera (MTMC) tracking systems.
“As MTMC methods solve larger and larger problems,” the paper reads, “it becomes increasingly important (i) to agree on straightforward performance measures that consistently report bottom-line tracker performance, both within and across cameras; (ii) to develop realistically large benchmark data sets for performance evaluation; and (iii) to compare system performance end-to-end.”
In the paper, the researchers presented a data set they had compiled, consisting of more than two million image frames of around 2,000 students from eight cameras placed around campus. The data set was known as the DukeMTMC. The study was funded by the U.S. Army Research Office and the National Science Foundation.
The 85 minutes of video from each camera was captured on one day in March 2014, when researchers knew students would be transitioning between classes.
Duke, meanwhile, does not plan to reopen the site hosting the DukeMTMC data set for facial recognition training, and acknowledged that its previous availability violated University policy, according to the The Chronicle.
Duke Vice President for Public Affairs and Government Relations Michael Schoenfeld wrote in an email to The Chronicle that media reports about the data sets use by the industry, and the lack of consent from those pictured, led to an investigation by the school’s Institutional Review Board.
The Board found that the data set was “neither collected nor made available to the public consistent with the terms of the study that had been approved by the Institutional Review Board.”
The data set consists of more than two million images of two thousand students, collected in 85 minutes of video from eight cameras placed around the campus in 2014.
The use of American resources to train technology by Chinese companies has recently come under scrutiny, and The Chronicle reports that just under half of the worldwide verified citations of the DukeMTMC data set are from Chinese researchers.
The post Facial Recognition Software Developed at Duke Linked to Chinese Surveillance Programs appeared first on Security Sales & Integration.