MSAAN

Multi-Scale Attentive Aggregation for LiDAR Point Cloud Segmentation

Introduction

We proposed a multi-scale attentive aggregation network for semantic segmentation of LiDAR point cloud. We test our network on two datasets which are CSPC dataset and Tonronto3D dataset.

Dependencies

Run

  1. Setup python environment
    2. 
     
    conda create -n randlanet python=3.5
    
     
    source activate randlanet
    
     
    pip install -r helper_requirements.txt
    
     
    sh compile_op.sh
  2. Data
    3. 
     
    We put the preprocessed data in /data/cspc and /data/toronto3d.
  3. Train
    4. 
     
    cspc:python -B main_CSPC.py --gpu 0 --mode train --test_area 5
    
         
    toronto3d:python -B main_toronto3d.py --gpu 0 --mode train --test_area 5
  4. Test
    5. 
     
    cspc:python -B main_CSPC.py --gpu 0 --mode test --test_area 5
    
         
    toronto3d:python -B main_toronto3d.py --gpu 0 --mode test --test_area 5

Download

Click HERE to Download