Scan Point Cloud Seg Dataset

1 Running Settings

1. Create your conda environment and activate it.

   conda create -n <your_env_name> python=3.6 --yes
   conda activate <your_env_name>

2. Install python package independency.

   pip install -r requirements.txt

2 Processing Procedure

Branch 1: input points file (.xyz, only xyz) and textured mesh file (.obj), output segmentation label file (.xyz/.npy, xyzrgbl).

1. Label foreground points from raw point cloud with tool software such as MeshLab.
2. Extract background points and generate point cloud segmentation label file by running generate_seg_label.py.
3. Generate point cloud segmentation label file with texture information by running generate_color_points.py.
4. 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)

3 Demo Dataset

We offer a demo ipad scanned dataset. You can download it by:

• Google Drive

• Baidu YunPan (extract code: ih8g)

4 Display

5 Point Cloud Segmentation Test

Please refer to the directory point-transformer-ipadscan.

6 To-DO List

1. update point-transformer-ipadscan README file.

7 Renference

[1] Python Package: mesh-to-sdf

[2] Point Transformer