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

Table 2

Pixel accuracy and segmentation time of UPerNet with different component analysis modules (CAMs) on ADE20K component segmentation task.

	Backbone	Backbone depth	Decoder	Comp. Analysis model	Comp. Segmentation accuracy (%)	Segmentation time (ms)
1	ResNet	50	PPM + FPN	—	48.30	483
2	ResNet	101	PPM + FPN	—	48.71	598
3	ResNet	152	PPM + FPN	—	48.89	721
4	ResNet	50	PPM + FPN + CAM	[29]	53.62	490
5	ResNet	101	PPM + FPN + CAM	[29]	53.96	604
6	ResNet	152	PPM + FPN + CAM	[29]	54.18	726
7	ResNet	50	PPM + FPN + CAM		54.03	492
8	ResNet	50	PPM + FPN + CAM		55.13	486
9	ResNet	50	PPM + FPN + CAM		55.62	496