torch-semantic-seg To dos 1. create datasets 1. cityscapes style dataset pipeline wrappers 2. create data pipelines 1. Load Image and annotation 2. Simple augmentations (FlipLR, FlipUD) 3. Dataset concat model wrappers 3. create models 1. Encoders (Backbones) 2. Decoders (Neck) 3. Segmentation Head 4. config 5. create loss, optimizer, schedule, save model 6. training & evaluation 1. Semi-Supervised Training 7. distributed training Referneces [1] PyTorch - WRITING CUSTOM DATASETS, DATALOADERS AND TRANSFORMS