Voxelnet ROS Implementation

---

VoxelNet Implementation codes from "https://github.com/qianguih/voxelnet"

envs:

• ubuntu 20.04
• ROS:noetic

sudo apt-get install python3.5
sudo apt install python3.5 python3.5-dev python3.5-venv

sudo apt-get install ros-noetic-jsk-recognition-msgs 
sudo apt-get install ros-noetic-jsk-rviz-plugins

Dependencies

• python3.5+
• TensorFlow (tested on 1.4)
• opencv
• shapely
• numba
• easydict
• ROS
• jsk package https://github.com/jsk-ros-pkg/jsk_recognition

Data Preparation

Download the test data: https://pan.baidu.com/s/1kxZxrjGHDmTt-9QRMd_kOA

unzip to data folder,Directory structure should be:

data
----lidar_2d
--------0000...1.npy
--------0000...2.npy
--------.......

Instructions

before run the code, you may need to install:

• clone this repository
• move voxelnet_ros folder to your catkin_ws
• catkin_make
• roscd voxelnet/script/
• python3 voxelnet_ros.py & python3 pub_kitti_point_cloud.py
  • unfortunately, rosrun won't work. because it's using Python 3.x instead of 2.x

ma upate:

• python3.5 voxelnet_ros.py & python3.5 pub_kitti_point_cloud.py

ROS Node

Rviz Animation

Future Work

• Retrain the model
• Still too slow, need more efficient implementation for the VoxelNet
• code clean