LaneKerb Detection

LaneKerb detection package is a real-time road feature detection method which includes the lane and kerb detection under various road scenarios and diverse weather conditions. It can provide road boundary information and lane-level information for autonomous vehicle navigation.

Fig. 3 Demo on NTU campus

Features

• Lable your data with Scalabel annotation tool.

• Data format conversion bdd-data needed for training LaneKerbNet deep neural network using your data labeled by Scalabel.

• A ROS node LaneKerbNetRos for lane and kerb detection using instance segmentation deep neural network.

• Able to handle multiple lanes and multiple kerbs on straight and curvy roads, at roundabout as well as on slope.

• Able to handle lane and kerb detection on rainy days.

Fig. 4 Scalabel lane and kerb lable

Lane and kerb groundtruth, class and instance visualization.

LaneKerb Detection Contents

• How to Train LaneKerbNet
  • 1. Installation
  • 2. Test model
  • 3. Train your own model
  • 4. Experiment
• LaneKerb Detection User Guide
  • 1. Dependencies
  • 2. Launch the node
  • 3. Demo on bus
• LaneKerbNet Algorithm
  • 1. Introduction
  • 2. Method
  • 3. Experiments

API REFERENCE

• Class Hierarchy
• File Hierarchy
• Full API
  • Namespaces
  • Classes and Structs
  • Functions
  • Variables