Source code for paper "Point Cloud Compression with Implicit Neural Representations: A Unified Framework".
The two softwares tmc3 and pc_error are already contained in this repo. Use the following command to change file permissions.
sudo chmod 777 tmc3 pc_error
Make sure that the configuration file config.yaml
is available in exp/geometry
. Then use the following commands for geometry compression.
python train.py --cloud_path=loot_vox10_1200.ply --exp_dir=exp/geometry
python test.py --cloud_path=loot_vox10_1200.ply --exp_dir=exp/geometry
The results can be found in exp/geometry/result.csv
.
Make sure that the configuration file config.yaml
is available in exp/attribute
. Then use the following commands for attribute compression.
python train.py --cloud_path=loot_vox10_1200.ply --geometry_path=exp/geometry/cloud.ply --exp_dir=exp/attribute
python test.py --cloud_path=loot_vox10_1200.ply --geometry_path=exp/geometry/cloud.ply --exp_dir=exp/attribute
The results can be found in exp/attribute/result.csv
.
This code allows compression using G-PCC, by directly calling the tmc3 software with predefined configurations.
Use the following command for geometry compression by G-PCC.
python baseline.py --cloud_path=loot_vox10_1200.ply --result_dir=gpcc --pqs=0.75
Use the following command for joint geometry and attribute compression by G-PCC.
python baseline.py --cloud_path=loot_vox10_1200.ply --result_dir=gpcc --encode_colors --pqs=0.75 --qp=34
The results can be found in gpcc/result.csv
.