Minimum Ncuts Example For 3D Point Cloud

Install

pip install -r requirements.txt 

Run

python run_ncuts.py ##runs on 2 chunks and visualized + stores output .pcd files to out_pcds/ folder

Files

• run_ncuts.ipynb & run_ncuts.py : main files to run and visualize some chunks, both files do the same
• normalized_cuts.py : contains main code for running ncuts
• point_cloud_utils.py : contains some helper functions for visualization and reprojecting points from downsampled pcd to original pointcloud

Params

This code is only using spatial distances, so the only relevant ncuts parameter is the ncuts_threshold T. Increasing the threshold inside config_spatial dict will lead to more recursive splits (finer grained/more instances).

Credits & Contributions

Laurenz Heidrich (@laurenzheidrich)