May I ask how to obtain semantic segmentation results? In the demo.py file you provided, it seems that only panoramic segmentation is available