Performance Comparison of Moving Target Recognition between Faster R-CNN and SSD
Hongyu Jin, Ming Gao Yan, Jinyu Lu, Lingyao Zhu, Kaili Wang, Shilei Bai · 2019 International Joint Conference on Information, Media and Engineering (IJCIME) · 2019
Target detection algorithms based on deep learning and convolutional neural networks (CNN), such as Faster Regionbased Convolutional Neural Network method (R-CNN) and Single Shot Multi-Box Detector (SSD), are superior to traditional algorithms in detection accuracy and speed. However, there is still a lack of detailed research on the different performance between different algorithms. This paper introduces the Faster R-CNN algorithm and SSD algorithm respectively, and implements the algorithm function based on the open source code. In order to evaluate the performance of these two algorithms, a large number of videos containing cars in different scenarios were used for testing. Experimental results show that the former is superior to the latter in recognition accuracy, while the latter is superior to the former in recognition accuracy.