PyTorch point cloud is a pytorch implementation collection of point cloud utils, models, and modules.
The Componets of this repository is as following.
examples
: include scripts to train or test implemented models.- Test results of models are shown on README.md (ex: examples/PointNet/README.md) in the model folder.
- Please look at a list in "Evaluated model" section when checking evaluated models.
- Example codes work by adding
TORCHPCP_DEBUG=true
to environment variable, even if you don't install torchpcp package using setup.py.
- Test results of models are shown on README.md (ex: examples/PointNet/README.md) in the model folder.
torchpcp
: include model, module, and utils.- Docstring guide of this repo follows numpy.
- Python3 (3.7.7)
- PyTorch (1.7.0)
- easydict
- Ninja (1.8.2)
- If you have not installed this software, Please show here.
- (optional) Hydra (1.0 or later)
- (optional) tensorboardX (2.1 or later)
python setup.py install
- Docker: nvidia/cuda:10.2-cudnn7-devel-ubuntu18.04
- GPU: RTX2080Ti x 1
- The following models were trained and tested in my environment.
- README.md of the following models shows the results of the model in implementations of this repository and papers.
- PointNet
- PointNet2ASIS (ASIS)
- DGCNN
- Model of this repository
- Paper:
- Original implementation:
- Model of this repository
- Paper:
- Original implementation:
- Model of this repository
- Paper:
- Original implementation:
- Model of this repository
- Paper:
- Original implementation:
- Model of this repository
- Paper:
- Original implementation:
- Model of this repository
- Paper:
- Original implementation: