Primitive Decomposition: Deep Learning Network 2D Natural Image Decomposition Easy dependency install from requirements.txt
conda create -n cv
conda activate cv
pip install -r requirements.txt
Basic structure:
Backbone of the network (to be worked on): backbone_training.py
Decomposition Network and primitive modules: network.py
encoder part for feature extraction: encoder.py
losses for quasi unsupervised task: losses.py
primitive geometry and distance field computation: field_geometry.py
Different notebooks for different datasets (to be worked on to split into python files and do one general training file)