Open AccessImproved Smoking Target Detection Algorithm Based On YOLOv3
Author(s) -
Wei Zhao,
Yun Zhu,
Qi Xuan Li,
Wei Cai
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1883/1/012052
Subject(s) - computer science , generalization , artificial intelligence , minimum bounding box , bounding overwatch , function (biology) , pattern recognition (psychology) , object detection , point (geometry) , image (mathematics) , algorithm , mathematics , mathematical analysis , geometry , evolutionary biology , biology
For the detection of smoking behavior, in order to make accurate state judgments, a smoking target detection algorithm based on the improvement of YOLOv3 is proposed. By introducing Mosaic data to enhance the background of the rich image, the batch_size is increased, and Mish is used as the activation function. To improve the generalization ability of the model, the DIOU_Loss function is used as the activation function to effectively use the information of the center point of the bounding box to improve the accuracy of target detection. Through self-made smoke detection data sets, experiments are carried out, and the experimental results show that the algorithm can accurately detect To smoking behavior, improve the MAP and AP of training.