Multiscale Point Cloud Geometry Compression

We apply an end-to-end learning framework to compress the 3D point cloud geometry (PCG) efficiently. Leveraging the sparsity nature of point cloud, we introduce the multiscale structure to represent native PCG compactly, offering the hierarchical reconstruction capability via progressive learnt re-sampling. Under this framework, we devise the sparse convolution-based autoencoder for feature analysis and aggregation. At the bottleneck layer, geometric occupancy information is losslessly encoded with a very small percentage of bits consumption, and corresponding feature attributes are lossy compressed.

News

2021.11.23 We proposed a better and unified PCGC framework based on PCGCv2, named SparsePCGC. It can support both lossless and lossy compression, as well as dense point clouds (e.g., 8iVFB) and sparse LiDAR point clouds (e.g., Ford). Here is the links: paper code
2021.7.28 We have simplified the code, and use torchac to replace tensorflow-compression for arithmetic coding in the updated version.
2021.2.25 We have updated MinkowskiEngine to v0.5. The bug on GPU is fixed. And the encoding and decoding runtime is reduced.
2021.1.1 Our paper has been accepted by DCC2021! [paper] [presentation]

Requirments

python3.7 or 3.8
cuda10.2 or 11.0
pytorch1.7 or 1.8
MinkowskiEngine 0.5 or higher (for sparse convolution)
torchac 0.9.3 (for arithmetic coding) https://github.com/fab-jul/torchac
tmc3 v12 (for lossless compression of downsampled point cloud coordinates) https://github.com/MPEGGroup/mpeg-pcc-tmc13

We recommend you to follow https://github.com/NVIDIA/MinkowskiEngine to setup the environment for sparse convolution.

Pretrained Models: https://box.nju.edu.cn/f/b2bc67b7baab404ea68b/ (or Baidu Netdisk link：https://pan.baidu.com/s/14kPV1ZJPWCsuGM0V6mWytg?pwd=pcgc)
Testdata: https://box.nju.edu.cn/f/30b96fc6332646c5980a/ (or Baidu Netdisk link: https://pan.baidu.com/s/1ccLR1fBrupIOeb2_0fPvQw?pwd=pcgc)
Training Dataset: https://box.nju.edu.cn/f/1ff01798d4fc47908ac8/

Usage

Testing

Please download the pretrained models and install tmc3 mentioned above first.

sudo chmod 777 tmc3 pc_error_d
python coder.py --filedir='longdress_vox10_1300.ply' --ckptdir='ckpts/r3_0.10bpp.pth' --scaling_factor=1.0 --rho=1.0 --res=1024
python test.py --filedir='longdress_vox10_1300.ply' --scaling_factor=1.0 --rho=1.0 --res=1024
python test.py --filedir='dancer_vox11_00000001.ply'--scaling_factor=1.0 --rho=1.0 --res=2048
python test.py --filedir='Staue_Klimt_vox12.ply' --scaling_factor=0.375 --rho=4.0 --res=4096
python test.py --filedir='House_without_roof_00057_vox12.ply' --scaling_factor=0.375 --rho=1.0 --res=4096

The testing rusults of 8iVFB can be found in ./results

Training

 python train.py --dataset='training_dataset_rootdir'

Authors

These files are provided by Nanjing University Vision Lab. And thanks for the help from Prof. Dandan Ding from Hangzhou Normal University and Prof. Zhu Li from University of Missouri at Kansas. Please contact us (mazhan@nju.edu.cn and wangjq@smail.nju.edu.cn) if you have any questions.

NJUVISION / PCGCv2