You are using the wrong function to perform semantic segmentation. segmentImage is the attribute of the instance segmentation class, not semantic segmentation. There are two options available for performing semantic segmentation, it is either you use pascal voc model that detects 20 common objects, read the tutorial from here https://github.com/ayoolaolafenwa/PixelLib/tree/master/Tutorials

or ade20k model that detects 150 objects, read the tutorial from here https://github.com/ayoolaolafenwa/PixelLib/blob/master/Tutorials/image_ade20k.md.