Laser & Optoelectronics Progress, Volume. 57, Issue 18, 181505(2020)
Improved Pneumonia Detection Algorithm Based on YOLOv3
Figures & Tables(7)
Fig. 2. Ordinary convolution and dilated convolution. (a) Dilation rate of 1; (b) dilation rate of 2
Fig. 4. Detection effect of different algorithms. (a) SSD algorithm; (b) YOLOv3 algorithm; (c) Faster-RCNN algorithm; (d) Multi branch YOLO algorithm
Table 1. Detection accuracy of different algorithms
View tableTable 1. Detection accuracy of different algorithms
Algorithm AP@0.4 AP@0.5 AP@0.6 mAP SSD[ 13 ]0.558 0.412 0.246 0.405 Faster-RCNN[ 7 ]0.573 0.426 0.271 0.423 YOLOv3[ 9 ]0.572 0.431 0.269 0.424 Multi branch YOLO 0.589 0.453 0.275 0.439
Table 2. Influence of dilation rate and maximum entropy on detection accuracy
View tableTable 2. Influence of dilation rate and maximum entropy on detection accuracy
Darknet 53 FPN Dilated convolution Maximum entropy mAP P P 0.424(YOLOv3) P P P 0.431 P P P P 0.439(ours)
Tools
Get Citation
Copy Citation Text
Shuhao Ma, Jubai An. Improved Pneumonia Detection Algorithm Based on YOLOv3[J]. Laser & Optoelectronics Progress, 2020, 57(18): 181505
Paper Information
Category: Machine Vision
Received: Nov. 19, 2019
Accepted: Feb. 14, 2020
Published Online: Sep. 2, 2020
The Author Email: Ma Shuhao (mashuhao1789@126.com), An Jubai (jubaian@dlmu.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20