Canaan announced its face mask detection model has been successfully deployed in the PaddlePi-K210 terminal core computing module that was jointly developed with Baidu. Canaan’s AI product developers further explained that the algorithm can detect whether the user is wearing a mask or get covered with any body part. Canaan said the Company will actively promote the open source of its face mask detection model and the development of related products.
According to the visuals of the model, we can directly see how the face mask detection model works. Upon capturing the user’s face by the camera, if the user is not wearing a mask or using any body part to cover up, the display frame is lighted in red and the user is warned; if the user wears the mask correctly, the display frame is lighted in green.
The main principle of the model is to use Mobilenet-YOLO for target detection and conduct subsequent judgement whether user is wearing mask. The model can screen up to 30 faces in one frame, which can meet the detection needs of face masks in densely-populated scenarios such as office building and bus/railway stations.