Facial recognition systems are not perfect, but they can still make mistakes, especially when the faces they evaluate are colored people. Researchers from the University of North Carolina now find that there are indeed many problems with the company's development of high-tech security systems. They show that some existing systems can be confused by their creation of VR-like, computer-rendered faces. In addition, they did a 3D model and they showed a security system on the phone, using only photos taken on social networks such as Facebook.
The team collected photos of 20 volunteer subjects from online resources. Then they created a 3D model of the volunteer's face, added some facial expressions and adjusted their eyes so they looked like they looked at the camera. In the case of them, they did not find any of the full face of the display object, they rebuilt the missing parts, even those areas shadows and textures. What makes this even more impressive is that some volunteers are security researchers and can only dug about three low-quality photos online.
Because the researchers' 3D models were shaded and even moved, they were able to fool 55%-85% of the safety systems they tested. Microsoft has a commercial solution that uses alternative hard hardware: Tobii's eye-tracking camera. However, a cost-effective addition of hardware and hardware vendors will need to decide whether there is sufficient demand and benefits for consumers to add dedicated components like infrared cameras or structured light projectors.
Fiber Cement Board Autoclave,Horizontal Autoclave,Autoclave For Fiber Cement Board,Fiber Cement Board Pressure Vessel
JIANGSU OLYMSPAN THERMAL ENERGY EQUIPMENT CO.,LTD , https://www.compositesautoclave.com