conda create --name apusmog python=3.8

conda activate apusmog

conda install pytorch=1.9.0 torchvision torchaudio cudatoolkit=10.2 -c pytorch

pip install h5py python-graph-core scipy PyYAML

cd third_party/lib_pointtransformer/pointops/
python setup.py install

cd third_party/pointnet2
python setup.py install

python main.py --config configs/apusmog_pu1k_pretrained.yaml

python main.py --config configs/apusmog_pu1k.yaml

cd evaluation
./run_me.sh

python compute_p2m.py --gt_dir ../data/PU1K/test/original_meshes/ --pred_dir ../checkpoints/apusmog_pu1k_pretrained/results/ --use_mp True
python evaluate_tf_cpu.py --gt_dir ../data/PU1K/test/input_2048/gt_8192/ --pred_dir ../checkpoints/apusmog_pu1k_pretrained/results/ --save_path ../checkpoints/apusmog_pu1k_pretrained/metrics --use_p2f

@inproceedings{delleva2022arbitrary,
    author = {Anthony Dell'Eva and Marco Orsingher and Massimo Bertozzi},
    title = {Arbitrary Point Cloud Upsampling with Spherical Mixture of Gaussians},
    booktitle = {International Conference on 3D Vision (3DV)},
    year = {2022}
}