Table 1.

Mask-RCNN network configuration.

Parameter	Value
GPU count	1
Images per GPU	2
Steps per epoch	1000
Validation step	50
Learning rate	0.001
Weight decay	0.0001
Back bone	ResNet101
FPN layers’ size	1024
RPN NMS threshold	0.7
Number of classes	1
Anchors per image	256
Detection minimum confidence	0.7