https://github.com/adioshun/Didi_challenge

1. 개요

자율주행 차 개발의 주요 Task중 하나는 주변 상황을 이해 하는 것이다.

본 Challenge를 통해 참가자는 LIDA데이터와 Camera데이터를 활용하여 주변 상황을 인지 하여야 한다.

1.2 데이터셋

awesome-vehicle-datasets

1.3 주요 링크

Getting Started Didi-Challenge

https://github.com/udacity/didi-competition/blob/master/docs/GettingStarted.md

1. 개요

Challenge 목표 : Detecting and locating obstacles in 3D space, 평가방법

2. Dataset

학습 데이터는 ROS bag 파일로 제공됨

  • 1,2의 차이 살펴 보기(1은 Training), bag파일은 1번에 있음

A bag contains the synchronized output of several ROS nodes.

vehicle bag

  • Camera video
  • Lidar point clouds
  • GPS/IMU measurements

Obstacle bags

  • Front RTK GPS
  • Back RTK GPS
[Tip] Convert /velodyne_packets to /velodyne_points ,출처

일부 데이터셋에는 용량문제로 Velodyne point cloud(published on the /velodyne_points ROS topic) 데이터가 없을수 있음

The LIDAR readings are represented in a compressed packet form (/velodyne_packets)

# Install the Velodyne package
sudo apt-get install ros-indigo-velodyne

# Run the conversion tool for the HDL-32E LIDAR unit that Udacity used to record the data
rosrun velodyne_pointcloud cloud_node _calibration:=/opt/ros/indigo/share/velodyne_pointcloud/params/32db.yaml

Now when you play a bag file with the /velodyne_packets topic, it will automatically get converted to a point cloud format and republished as /velodyne_points

3. 활용 소프트웨어

  • ROS : 데이터셋 추출 및 가공

  • RVIZ : 데이터 시각화

  • Autoware : camera/LIDAR calibration.


2. Solution

2.1 omgteam

  • omgteam: This repository is to provide visualization, calibration, detection ROS nodes.

2.2 hengcherkeng

2.3 Team Timelaps(markstrefford)

2.4 experiencor

포인트 클라우드가 아닌 이미지에 3D Bbox 적용하는 방법

2.5 Boston Team

2.6 MV3D_TF

https://github.com/adioshun/MV3D_TF

2.9 etc

3. References

3.1 Article/Blog

3.2 Tutorial

A. Ronny Restrep

4. Tools

4.1 ROS Viwer

4.2 Docker

4.3 변환 툴


(kitti_download)[https://gist.github.com/adioshun/0554effff45e4f16fe4db7eb1c4712cc)

results matching ""

    No results matching ""