This is the source code for our paper .

A Novel Semi-Supervised Method for Airborne LiDAR Point Cloud Classification

Introduction

This is the source code for our paper A Novel Semi-Supervised Method for Airborne LiDAR Point Cloud Classification. This repo only includes the code for conducting experiments using RandLA-Net as the backbone network. Our original repo using DANCE-Net as backbone network can be found at SemiALS-Net.

Installation

Please refer to official RandLA-Net.

Data Preparation:

python utils/data_prepare_isprs_large_w_height_mask.py

python utils/data_prepare_dfc.py

ISPRS dataset

training: python main_isprs.py --mode train --gpu 0

evaluation: python main_isprs.py --mode test --gpu 0

DFC 3D dataset

training: python main_dfc.py --mode train --gpu 0

evaluation: python main_dfc.py --mode test --gpu 0

Acknowledgements

Large Part of the code is borrowed from RandLA-Net.