DSSLIC: Deep Semantic Segmentation-based Layered Image Compression

Datasets

-Cityscapes https://www.cityscapes-dataset.com/ -ADE20K https://groups.csail.mit.edu/vision/datasets/ADE20K/ -VOC2012

Quickstart:

• create env

conda env create -f environment.yml
conda activate DSSLIC

• download & unpack ADE20K and run src/data/ade20k_utils (set path to dataset inside), to preprocess segmentations
• copy config.template.yml to config.yml
• set dataset path in config file
• run scripts from src directory (PYCHARM: right click src and "Mark directory as / Sources root")
• run scripts with project root as cwd (PYCHARM: run configuration settings)
• python source/train.py