-
Create your conda environment and activate it.
conda create -n <your_env_name> python=3.6 --yes conda activate <your_env_name>
-
Install python package independency.
pip install -r requirements.txt
Branch 1: input points file (.xyz, only xyz) and textured mesh file (.obj), output segmentation label file (.xyz/.npy, xyzrgbl).
- Label foreground points from raw point cloud with tool software such as MeshLab.
- Extract background points and generate point cloud segmentation label file by running
generate_seg_label.py
.- Generate point cloud segmentation label file with texture information by running
generate_color_points.py
.- Generate augmented train data and test data by running
generate_train_test_split_data.py
.
Branch 2: input points file (.ply, xyzrgb), output segmentation label file (.xyz/.npy, xyzrgbl).
TO-DO (refer to
ply_process.py
)
We offer a demo ipad scanned dataset. You can download it by:
Please refer to the directory point-transformer-ipadscan
.
- update point-transformer-ipadscan README file.