A Framework for Automatic Burn Image Segmentation and Burn Depth Diagnosis Using Deep Learning

Table 3

Results of different backbone networks on testing set.

	IOU	PA	DC
HRNetV2-C1 no-Aug	0.8308	0.9331	0.9076
HRNetV2-C1 Aug	0.8467	0.9403	0.9170
ResNet-50-dilated-PPM no-Aug	0.8136	0.9231	0.8972
ResNet-50-dilated-PPM Aug	0.8412	0.9374	0.9138
ResNet-50-UPerNet no-Aug	0.8147	0.9321	0.8979
ResNet-50-UPerNet Aug	0.8341	0.9375	0.9096
ResNet-101-dilated-PPM no-Aug	0.7864	0.9215	0.8804
ResNet-101-dilated-PPM Aug	0.7913	0.9128	0.8835
ResNet-101-UPerNet no-Aug	0.8162	0.9459	0.8988
ResNet-101-UPerNet Aug	0.8259	0.9172	0.9046