Marathon Athletes Number Recognition With Deep Learning

In the marathon competition,a large number of photos are taken for each athlete in the photo pushing service.Therefore,how to classify photos of specific athletes accurately and effectively has become the focus of attention.This paper proposes a compound deep neural network for marathon athletes number recognition to make classification more efficient and accurate.The proposed model is divided into three modules: image preprocessing module,text detection module,and text recognition module.Firstly,in the preprocessing module,this paper makes use of the You Only Look Once version 3(YOLOv3)to extract the athlete areas,and sets the detection threshold and similarity threshold to reduce unnecessary detection and improve the overall efficiency of the model.Secondly,in the text detection module,this paper combines the efficient text detector An Efficient and Accuracy Scene Text detection pipeline(EAST)to detect the texts in the athlete areas.Besides,in the text recognition module,this paper uses transfer learning to fine-tune the CRNN to make it suitable for the marathon athletes number recognition.In addition,this paper designs an effective tree filtering algorithm to avoid the interference caused by the text detection module,and improve the detection accuracy of the model.Finally,to solve the problem of the lack of data related to the marathon athlete number recognition,this paper collects and produces two datasets,MARD and MANUMD,for the training and testing of the model.In this paper,relevant comparative experiments are designed for each module,and the overall efficiency of the model is analyzed.The experimental results show that the model is feasible and effective and can be applied to the image classification of marathon photo pushing services.