Autoware YOLOv3 and PointPillars

Author: Kemiao HUANG

Last update: 2019.9.4

1. Install Autoware.AI from Source

See installation tutorial on https://gitlab.com/autowarefoundation/autoware.ai/autoware/wikis/Source-Build for master branch build.

wget -O autoware.ai.repos "https://gitlab.com/autowarefoundation/autoware.ai/autoware/raw/master/autoware.ai.repos?inline=false"

2. Download ROSBAG

Download ROSBAG files from server.

sftp [email protected] 
# Password is not shown here. You may use other rosbag from web or record you rosbug by yourselves. If you want to use our dataset, please contact me.
cd 2019-03-24/
get 2019-03-24-16-29_X.bag

Extract topics from bags to fit Autoware format.

python2.7 extract_topics.py in.bag out.bag

3. Download Weights

Use the pre-trained weights for pointpillars and yolov3 or train your own models. For more details, please see README.md in Autoware perception modules.

4. Run Autoware

Run ROS and rosbag.

roscore

Source setup file.

source $HOME/autoware.ai/install/setup.bash

Change the settings in my_launch.launch file and run it.

roslaunch my_launch.launch

Run Rviz with the configuration file.

rosrun rviz rviz -d sustech-2019-03-24.rviz

Run ROSBAG and see the test result in rviz.

rosbag play XXX.bag