search

loicland / superpoint_graph

Large-scale Point Cloud Semantic Segmentation with Superpoint Graphs
MIT License
764 stars 214 forks source link

The method mentioned in your paper #23

Closed LZDSJTU closed 6 years ago

LZDSJTU commented 6 years ago

In your paper: WEAKLY SUPERVISED SEGMENTATION-AIDED CLASSIFICATION OF URBAN SCENES FROM3D LIDAR POINT CLOUDS You mentioned that you implemented two other methods: pointwise and CRF regularization. Could you share the code? So I can compare the two methods. Thank you!

loicland commented 6 years ago

Hi,

the pointwise method was simply a random forest classifier trained from the same geometric features as the one computed in this repository by libply_c.compute_geof.

The CRF methods are adapted from this repository (matlab wrappers only).

FYI these methods will massively underperform compared to superpoint graph or any modern approach of point cloud segmentation.

loicland commented 6 years ago

closing this this is not related to this repo.