This project is an implementation of Li, L., Yang, F., Zhu, H., Li, D., Li, Y., & Tang, L. (2017). An Improved RANSAC for 3D Point Cloud Plane Segmentation Based on Normal Distribution Transformation Cells. Remote Sensing, 9(5), 433. doi:10.3390/rs9050433 as a part of the second year coursework.
It does not work with Google Colab, so use Jupiter Notebook or just something else.
The results completely depend on the pointcloud you have. The one I had tested on had a mirror reflection and bad light spot. Somewhere it was chunky, but in general it worked good with the less noisy parts.
Pointcloud I had:
Meshes I had got before fine-tuning and piece combining
After the piece combining. Orange mesh is a celling, green one is a floor.
One more 3D picture of the mesh.
