urban-scene-segmentation

Semantic Segmentation using U-Net and DeepLabV3+ in PyTorch with pretrained encoders, data augmentation, Dice + CrossEntropy loss, quantitative evaluation, and error analysis on the CamVid dataset.