TABLE 5.
Precision and speed of various models.
Method |
R-FCN |
Faster R-CNN |
Faster DR-IACNN |
|||||
Backbone | ResNet 50 | ZF | VGG 16 | GoogLeNet | ResNet 50 | ResNet 34 | ResNet 18 | INSE-ResNet (our work) |
mAP (%) | 74.8 | 66.9 | 67.5 | 64.3 | 68.6 | 71.5 | 69.7 | 81.1 |
Speed (FPS) | 15.75 | 16.08 | 15.85 | 18.65 | 7.11 | 10.33 | 13.94 | 15.01 |
The meaning of the “bold” is the best experimental results.