Table 2.

Detection and segmentation results of scar by using Mask region-based convolutional neural network (R-CNN) with feature proposal network (FPN) and various backbones (ResNet50, ResNet10, ResNeSt50, and ResNeSt101): mean average precision (mAP) and mean average recall (mAR).

Backbone	${\mathit{mAP}}^{\mathit{bbox}}$	${\mathit{mAR}}^{\mathit{bbox}}$	${\mathit{mAP}}^{\mathit{mask}}$	${\mathit{mAR}}^{\mathit{mask}}$	Time (s)
ResNet 50	0.598	0.666	0.619	0.672	0.05
ResNet 101	0.620	0.680	0.631	0.677	0.07
ResNeSt 50	0.564	0.641	0.613	0.659	0.07
ResNest 101	0.597	0.672	0.587	0.645	0.09