This repository provides a PyTorch implementation for semantic segmentation and scene parsing, featuring state-of-the-art models including PSPNet and PSANet that achieved first place in ImageNet Scene Parsing Challenge 2016, LSUN Semantic Segmentation Challenge 2017, and WAD Drivable Area Segmentation Challenge 2018. The codebase supports training and testing on datasets such as ADE20K, PASCAL VOC 2012, and Cityscapes, using ResNet50/101/152 as backbone architectures with multi-GPU distributed training support.