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

Figure 10

CITYSCAPES instance-level semantic labeling by UPerNet.