Personnel dress code detection algorithm based on convolutional neural network cascade

Abstract

In order to solve the problems of the lack of a unified personnel management system and the difficulty of manual supervision in the management of movie theater enterprises, this paper proposes a personnel dress code detection algorithm based on convolutional neural network cascade. The algorithm uses two Retinanet network models to cascade to detect target objects. The first network model detects large target objects, and the second network model performs secondary detection on the detected personnel target objects to detect the personnel's clothing target objects. At the same time, the deep separable convolutional layer is used to replace the conventional convolutional layer in the backbone network of Retinanet network model, which reduces the number of parameters of the network model and improves the detection speed of the algorithm. Finally, an improved regression loss function LIoU is proposed to further improve the detection accuracy of the algorithm. The experimental results show that: on the collected data set, the average detection accuracy (mAP) of the algorithm in this paper reaches 56.9%, and the detection time is 0.083 seconds per frame. And at the same detection speed, the average detection accuracy is improved by 4.9% compared to using a single network model.