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

Table 3

Performances of different methods on CITYSCAPES instance-level semantic labeling task.

Method	AP (%)	(%)	Segmentation time (ms)
SegNet	29.5	55.6	—
Mask R-CNN	32.0	58.1	—
UperNet	32.0	57.3	447
UperNet + CAM	36.5	62.2	451

CAM: Component Analysis Module.