This is the implementation of paper "point cloud colorization based on densely annotated 3d shape dataset".

Sample Results:

Download prepared training data from this link and decompress it into the data directory.

Just run: python train.py
Hyperparameters are saved in base_config.ini and could be tuned. Each time you train the model, a copy of the ini file is saved in train_results/training time/. The trained model parameters are saved in train_results/training time/category_name/model/; the intermediate colorization results of training data and test data are saved in train_results/training time/category_name/logs/.