ROS Packages to be used with LGSVL Simulator Micromobility version
This repository contains 4 ROS packages used for various purposes in the project using LGSVL Automotive Simulator. Place these packages into your ROS workspace and run catkin_make to build the packages.
This package provides a ROS node for data collection from the LGSVL Simulator. This currently supports collecting camera images, depth sensor images, LiDAR point clouds and 2D, 3D ground truths at synchronized time intervals.
- YOLO3D Style :
x, y, width, height, label - YOLOv2/YOLOv3 Style :
x_min. y_min, x_max, y_max, label
roslaunch lgsvl_data_collector data_collector_launcher1.launch
To save LiDAR pointclouds, make sure to run pointcloud_to_pcd node in the pcl_ros package in a separate terminal using the following command
rosrun pcl_ros pointcloud_to_pcd input:=/sync_pcl2 _prefix:="PATH_TO_SAVE_TO" _binary:=false
This package uses a trained YOLOv3 model to run inferences on incoming camera images from the LGSVL simulator. It outputs the results in two separate ways:
- In a separate window with detection boxes.
- Directly into the simulator by publishing to Autoware specific topics.
Trained weights can be downloaded from here. Add the path of this file to yolo.py file to load these weights while running inference.
If you would like to train on your own collected dataset, follow the instructions on this YOLOv3 repository.
roslaunch lgsvl_mm_perception mm_2d_perception_launcher.launch
This package has been provided by LGSVL and is cloned from their original repository here.
This package provides the messages that can be published to the LGSVL Simulator to display detections directly into the simulator camera window.
- Deepak Talwar - (https://github.com/deepaktalwardt)
- Seung Won Lee - (https://github.com/swdev1202)