Skip to main content
. 2023 Jan 21;23(3):1256. doi: 10.3390/s23031256

Table 2.

Comparison results of different models.

Model Precision (%) Recall (%) F1 (%) mAP (%) FPS Model Size (M)
Faster R-CNN [11] 59.74 81.78 69.50 76.53 10.45 108.64
CenterNet [16] 96.34 76.56 85.00 90.54 44.14 124.61
SSD [17] 91.31 77.11 83.50 88.53 33.86 100.27
EffificientDet-D2 [29] 96.51 60.10 72.00 76.92 15.64 31.2
Ghost-YOLOv4 [43] 90.77 84.07 87.00 89.83 36.78 43.6
YOLOv4-Tiny [42] 86.12 77.10 81.50 81.26 118.52 22.42
YOLOv4 [15] 91.80 85.30 88.50 92.46 23.02 243.92
Improved YOLOv4 92.78 86.35 90.00 92.98 43.24 41.88