YOLO-R Based Mask Recognition System with Winograd Convolution Acceleration Chip

ABSTRACT

In this paper, we proposed 'YOLO-R based mask recognition system with Winograd convolution acceleration chip.' In order to prevent the spread of COVID-19, we proposed a mask recognition system based on YOLO-R that adopted the CSP backbone architecture and used prediction refinement to predict, optimize, and obtain high recognition results. In addition, we designed the Winograd convolution acceleration chip to accelerate the operation of the YOLO-R neural network, so that the system can achieve real-time computing. © 2022 IEEE.