VoxelCNN is an order-aware generative model for building houses in Minecraft. This codebase is a PyTorch implementation of the training and evaluation pipeline for VoxelCNN.
VoxelCNN is trained and evaluated with the 3D-Craft dataset.
For more details, please refer to the ICCV 19' paper Order-Aware Generative Modeling Using the 3D-Craft Dataset.
Python version >= 3.7 is required.
git clone https://github.com/facebookresearch/VoxelCNN
cd VoxelCNN
pip install virtualenv
virtualenv voxelcnn_venv
source voxelcnn_venv/bin/activate
Please follow the official installation guide to install PyTorch version >= 1.3.
pip install numpy requests tqdm
python -m unittest discover -s test -v
The 3D-Craft dataset will be downloaded automatically when launching the training for the first time.
Run the fast training (fewer epochs, slightly worse results):
python main.py --num_epochs 3 --step_size 1 --save_dir /path/to/save/log/and/checkpoints
Example final test results for the fast training:
acc@1: 0.622 acc@5: 0.760 acc@10: 0.788
cca_10%: 13.374 cca_25%: 11.115 cca_50%: 12.546 cca_75%: 12.564 cca_90%: 7.632 cca_avg: 11.446
mtc: 131.411 mtc_normed: 0.241
Run the full training:
python main.py --save_dir /path/to/save/log/and/checkpoints
Example final test results for the full training:
acc@1: 0.640 acc@5: 0.778 acc@10: 0.806
cca_10%: 13.630 cca_25%: 12.223 cca_50%: 13.168 cca_75%: 13.047 cca_90%: 7.571 cca_avg: 11.928
mtc: 121.753 mtc_normed: 0.223
VoxelCNN is released under the CC-BY-NC 4.0 license.
@inproceedings{zchen2019,
title = {Order-Aware Generative Modeling Using the 3D-Craft Dataset},
author = {Chen, Zhuoyuan and Guo, Demi and Xiao, Tong and Xie, Saining and Chen, Xinlei and Yu, Haonan and Gray, Jonathan and Srinet, Kavya and Fan, Haoqi and Ma, Jerry and Qi, Charles R and Tulsiani, Shubham and Szlam, Arthur and Zitnick, C. Lawrence},
booktitle = {ICCV},
year = {2019},
}