This is an implementation of paper "Occupancy Networks - Learning 3D Reconstruction in Function Space" that is different from the official implementation. It includes a CRF-RNN module and a different training, validation and testing code.

1. Install Anaconda3.

2. Run the following commands to create conda environment and install all dependencies:

username@PC:~$ conda env create -f environment.yml
username@PC:~$ conda activate onet-crf

We follow the same data preparation described here.

In order to train and validate, cfg['mode'] must be 'train' in the main. Then:

username@PC:~$ python main.py.

To test it, cfg['mode'] must be 'test'. Then:

username@PC:~$ python main.py.

To evaluate the model, cfg['mode'] must be 'eval'. Then:

username@PC:~$ python main.py.

The output files and logs will be saved in cfg['out']['out_dir'].