Design and Implementation of Remote DeepFace Model Face Recognition System Based on sbRIO FPGA Platform and NB-IOT Module

As one of the main research directions in the field of identity recognition, face recognition plays an important role in fast and accurate applications. Today's face recognition technology has greatly improved the speed and precision under the support of deep learning algorithms, but it relies more on the huge processing. In the conventional mode, the high-definition picture needs to be transmitted back to the PC for processing, but the low-capacity, low-bandwidth, low-processor scene face recognition problem is not solved. Considering such extreme applications, we combine embedded FPGA technology with low-power narrow-band communication NB-IOT module [1] to form a narrow-bandwidth application framework. Using the DNN (Deep Neural Network) based with graphic FPGA programming, the front face recognition of the DeepFace model and the extraction of the 7-layer DNN convolution result are performed on the Zynq FPGA chip of sbRIO [2].Then through the remote transmission of NB-IOT, the classification data is sent back to the local server for comparison through the CoAP protocol [5] of the IOT operator, and the face recognition task can be completed.