Nuscenes Visualization with ROS-Rviz

This package aims to visualize nuscenes dataset in rviz environment.

We fully export nuscenes data to standard ROS topics.

Visualizing for Nuscenes in ROS

# modify path in launch file as necessary
roslaunch nuscene_visualize visualize_launch.launch 

Key features

• Nuscenes keyframes support.
• Six cameras.
• LiDARs.
• Ground truth bounding boxes.
• Pose and TF-trees (world, odom, base_link and sensors).
• GUI control & ROS topic control.
• Nuscenes sweeps support. (the images / lidar not synchronized well)
• RADARs.

Setup

Install nuscenes-devkit

pip3 install nuscenes-devkit

Install ROS, tested on Ubuntu 20.04, ROS noetic. (noted that the code was first developed on 18.04 with a mix of Python2/Python3). If you want to use this code in Ubuntu 18.04, please try:

## Checkout to the version with partial python2 support. Codes with nuScenes will run in python3 and codes with tf trees will be in python2. In new version, every thing is in Python3.
git checkout 2bc4abbcb5d8fca5e6b7df57012c6c5a505bab2d

## Enable rospy in Python3 for ROS melodic (this should not affect Python2), but it does not enable tf in Python3.
sudo apt-get install python3-catkin-pkg-modules
sudo apt-get install python3-rospkg-modules

Clone this repo into a ROS workspace/src and run

catkin_make
source devel/setup.bash

under the workspace folder.

Default Topics

Publish

topic name	Description
/nuscenes/{channel}/pose	sensor coordinates w.r.t. to base_link
/nuscenes/{channel}/image	image.
/nuscenes/{channel}/camera_info	intrinsic parameters.
/nuscenes/ego_pose	Relative pose base_link w.r.t to the world.
/nuscenes/LIDAR_TOP/data	Point cloud data.
/nuscenes/bboxes	MarkerArray visualizing bounding boxes.