These do not output raw images, the university’s webpage says, but perform their own computations, delivering “high-speed and low-power consumption” to enable “new embedded-vision applications in areas such as robotics, VR [virtual reality], automotive, toys [and] surveillance”.