Skip to main content
. 2022 Aug 4;13:889853. doi: 10.3389/fpls.2022.889853

Table 2.

Evaluation metrics of different trained models [ResNet50, ResNet101, and ResNet50 + Separable Convolution (SepConv)].

Network configuration Test plants
a b c d e f g Mean Inference time (s)
Average precision (AP) of plants mAP

  • ResNet50

 1.0 0.875 0.75 0.923 0.899 0.857 1.0 0.8672 ~1.92

  • ResNet101

 0.699 0.75 0.75 0.769 0.777 0.857 1.0 0.8036 ~2.42

  • ResNet50 + SepConv

 1.0 1.0 0.75 0.923 1.0 0.714 1.0 0.9045 ~1.75
F1 score of plants F1

  • ResNet50

 0.681 0.64 0.585 0.65 0.645 0.636 0.684 0.639

  • ResNet101

 0.614 0.597 0.585 0.591 0.614 0.636 0.684 0.619

  • ResNet50 + SepConv

 0.68 0.682 0.585 0.65 0.681 0.589 0.718 0.652
MCC of plants MCC

  • ResNet50

 0.888 0.904 0.857 0.848 0.89 0.933 0.943 0.88

  • ResNet101

 0.799 0.844 0.852 0.794 0.924 0.914 0.929 0.86

  • ResNet50 + SepConv

 0.904 0.923 0.866 0.88 0.923 0.896 0.914 0.89