• Laser & Optoelectronics Progress
  • Vol. 57, Issue 12, 121502 (2020)
Qiong Zhao1、2, Baoqing Li1、*, and Tangwei Li1、2
Author Affiliations
  • 1Key Laboratory of Science and Technology on Microsystem, Shanghai Institute of Microsystem and Information Technology, Chinese Academy of Sciences, Shanghai 201800, China
  • 2University of Chinese Academy of Sciences, Beijing 100049, China
  • show less
    DOI: 10.3788/LOP57.121502 Cite this Article Set citation alerts
    Qiong Zhao, Baoqing Li, Tangwei Li. Target Detection Algorithm Based on Improved YOLO v3[J]. Laser & Optoelectronics Progress, 2020, 57(12): 121502 Copy Citation Text show less
    Residual layer structure
    Fig. 1. Residual layer structure
    Darknet53 structural parameters
    Fig. 2. Darknet53 structural parameters
    Improved network structure
    Fig. 3. Improved network structure
    Network structure parameters of increase module
    Fig. 4. Network structure parameters of increase module
    Loss function training curve
    Fig. 5. Loss function training curve
    Model diagram of different schemes. (a) 1st kind; (b) 2nd kind
    Fig. 6. Model diagram of different schemes. (a) 1st kind; (b) 2nd kind
    Experimental renderings on test dataset
    Fig. 7. Experimental renderings on test dataset
    AlgorithmNetworkDatamAP /%
    Faster R-CNNVGGVOC2007+201273.2
    Faster R-CNNResidual-101VOC2007+201276.4
    R-FCNResidual-101VOC2007+201280.5
    SSD321Residual-101VOC2007+201277.1
    SSD500Residual-101VOC2007+201280.6
    YOLO v2_416Darknet19VOC2007+201276.8
    YOLO v3_416Darknet53VOC2007+201279.4
    OursDarknet53VOC2007+201280.2
    Table 1. Test results of various algorithms on VOC2007 dataset
    DatasetmAP /%
    BeforeAfter
    VOC200780.2080.47
    Table 2. Impact of data enhancement on model accuracy
    ModelSolution oneSolution two
    Time /h142134
    Table 3. Influence of different models on model training time
    Qiong Zhao, Baoqing Li, Tangwei Li. Target Detection Algorithm Based on Improved YOLO v3[J]. Laser & Optoelectronics Progress, 2020, 57(12): 121502
    Download Citation