Skip to content
Try to reproduce results of ECCV paper but far from it. Don't star, this is not my active repo!
Branch: master
Clone or download
Latest commit 3f0e581 Nov 23, 2018
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
cmake_modules first commit Aug 17, 2018
include
mesh
script first commit Aug 17, 2018
src first commit Aug 17, 2018
.gitignore
CMakeLists.txt first commit Aug 17, 2018
Readme.md Update Readme.md Nov 23, 2018
package.xml

Readme.md

Stereo vision based obstacle mapping

** Don't star, this is not my active repo!** Currently it is under a private repository.

Original plan is to merge the two inputs of obstacle bounding boxes and generate a optimized obstacle cubicle map.

First try to reproduce the result of paper: Stereo Vision-based Semantic 3D Object and Ego-motion Tracking for Autonomous Driving of ECCV 2018.

@article{li2018stereo,
  title={Stereo Vision-based Semantic 3D Object and Ego-motion Tracking for Autonomous Driving},
  author={Li, Peiliang and Qin, Tong and Shen, Shaojie},
  journal={arXiv preprint arXiv:1807.02062},
  year={2018}
}

The semantic_mapping_node subscribes the obstacle map information obstacle_detection::MapInfo from two sets of cameras and pose information geometry_msgs::PoseStamped /ugv_slam_node/posestamped from wide camera.

Prerequisite

The package is based on Ubuntu 16.04 with ROS Kinetics. Additionally you should install:

  1. OPENCV 3.1+

  2. Eigen 3.2+

  3. Ceres Solver

  4. Install VSLAM and object detection repositories in <YOUR_CATKIN_WORKSPACE>:

Installation

    cd catkin_ws/src

    git clone git@github.com:zhanghanduo/stereo_semantic_mapping.git

    cd ..

    catkin_make

Demo

    roslaunch semantic_mapping demo.launch

RoadMap

  • [] Early Development - Mapping without optimization
  • [] Object Tracking
  • [] Sparse Feature Observation
  • [] Semantic 3D Object Measurement
  • [] Point Cloud Alignment
  • [] Joint Optimization
  • [] Evaluation
You can’t perform that action at this time.