Implements the Fast-SCNN architecture for semantic segmentation, enabling pixel-level classification of images into semantic categories. The model is designed for fast inference and supports training on datasets like Cityscapes. Includes training, evaluation, and demo scripts for the segmentation task.