This study uses convolutional neural network algorithm to implement a mask recognition system based on RISC-V architecture embedded platform. By comparing the current deep learning algorithms, we choose to use TensorFlow platform, YOLO algorithm and mask recognition model, and train them with appropriate amount of sparsity to reduce the model size so that they can run on the embedded platform. Then Dlib is adopted to train the feature point detection model to extract and save the feature values of the face, which can achieve the goal of recognizing faces. The LFW dataset is used as the experimental sample, and the experiment proves that the method can achieve high accuracy.