计算智能和神经科学
研究文章

Multibranch对象检测方法对交通场景

表4

比较各种方法召回率(%)。
模型 阈值
0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65
RCNN [4] 76.23 74.55 71.29 68.78 65.46 61.12 59.29 57.04 54.35 49.47 46.78 42.04
快RCNN [9] 82.92 80.47 78.05 74.11 71.38 68.61 66.29 63.33 61.74 57.61 54.29 50.33
SSD (15] 79.74 77.07 74.14 72.97 70.61 68.14 66.24 63.39 61.27 58.14 55.24 51.39
面具RCNN [13] 80.65 78.08 76.54 75.15 72.09 68.98 65.66 64.21 63.37 60.25 56.45 53.10
SINet [2] 86.56 85.55 83.03 82.04 78.08 74.00 70.23 66.49 63.30 61.72 60.28 59.26
YOLOv3 [17] 85.94 82.07 80.34 78.30 75.35 73.82 71.32 67.20 65.14 61.29 59.32 56.73
MBNet网 88.25 86.31 83.45 80.67 77.94 74.19 71.71 67.83 64.20 62.19 60.81 58.33