Researchers develop AI method for movement identification and tracking without facial recognition

A team of Portuguese researchers have developed a way to identify and track individual animals with artificial intelligence but without facial recognition, which could eventually be applied to public surveillance of humans, Defense One reports.

The researchers used a convolutional neural network (CNN) to create idtracker.ai, which can recognize individual zebrafish and flies by their movements, with 99 percent accuracy for both. CNNs are commonly used in facial biometrics, and NIST recently singled them out as the advance most responsible for the dramatic improvement of the technology’s accuracy over the past five years. According to the researchers, idtracker.ai is “species agnostic,” so will work with people or any other kind of animal.

Microsoft called for government regulation of facial recognition in July of last year, saying it raises issues about privacy and other fundamental human rights. SensibleVision CEO George Brostoff responded in an email to Biometric Update that the use case, ownership, and storage of biometric data is what should be considered for regulation, rather than facial biometrics as such.

Gait recognition, which like idtracker.ai identifies individuals by the way they move, continued to advance in 2018, capped by Chinese startup Watrix AI raising roughly $14 million in early funding and launching its gait biometrics solution in October.

Article Topics

artificial intelligence  |  identity verification  |  neural networks  |  surveillance