Semantic Segmentation under a Complex Background for Machine Vision Detection Based on Modified UPerNet with Component Analysis Modules

Table 5

Pixel accuracy and segmentation time of UPerNet with different component analysis modules (CAM) for CNY anticounterfeit features via vision-based detection.

	Backbone	Depth	Decoder	Component analysis module	(%)	(%)	(ms)
1	ResNet	50	PPM + FPN	—	88.50	85.3	483
2	ResNet	50	PPM + FPN + CAM	[29]	90.38	96.1	490
3	ResNet	50	PPM + FPN + CAM		95.29	100	496