LaneKerb Detection User Guide

Contents

This project is implemented for lane and kerb detection. Input and output parameters can be config at launch file.

1. Dependencies

Install the dependencies:

pip install -r requirements.txt

Then you need to install the ROS kinetic.

That’s all.

In you catkin workspace src dir, clone the project:

cd catkin_ws/src/
git clone git@gitlab.com:ugv_stereo/lanekerbnetros.git

Download the trained weights from lanekerbnet/model/tusimple_lanenet.

Copy these checkpoints files to the folder lanekerbnetros/model/tusimple_lanenet.

Change the weight_path in lanenet.launch file.

Set the output image folder save_dir in lanenet.launch file.

Build and source (in the workspace)

cd ..
catkin_make
sorce devel/setup.bash

You may need to change the mode of the python script, as follow:

cd lanekerbnetros/scripts/
sudo chmod +x lanenet_node.py

Open one terminal and run

roscore

Open a new terminal and play one of you rosbag which contains the images

rosbag play --pause -l '/media/zh/Data/Dataset/rosbag/201905demo/stereo_2019-06-04-10-02-02_soonlee_loop.bag'

Subscirbe to /long/left/image_raw image topic by roslaunch

roslaunch stereo_viewer capture.launch

Then you can launch the lane detector node with

roslaunch lane_detector lanenet.launch

It’s realtime(30fps) in my GTX1080.

I run the ros node on the collected rosbag data and make it a video. You may catch a glimpse of it bellow.