【泡泡汇总】最强 SLAM Datasets 合辑

对于许多无人驾驶或者 SLAM 初学者来说，搞一个真车或者选择一款良好的设备来进行算法测试通常是个很头大的问题，且还需要解决很多诸如传感器之间标定不好，多传感器时间未做同步，没有真正的 groundtruth 来检验算法的好坏等等问题。

而跑 KITTI 便是解决这些问题的一种行之有效的方法，但如果我们想要更多的数据呢？本文来源于 GitHub 仓库 youngguncho/awesome-slam-datasets，整理出了几乎所有提供姿位姿和地图信息的各种 SLAM 数据集。在获得授权我们对其关键词进行了简要的翻译。可以直接访问我们的 GitHub repo。

https://github.com/PaoPaoRobot/awesome-slam-datasets 

值得补充的是，原文尚未提及 Apollo Scape 数据集和 ICRA 2019 的 Dataset Generation and Benchmarking of SLAM Algorithms for Robotics and VR/AR workshop 中的相关数据集（如The UZH-FPV Drone Racing Dataset 无人机快速竞赛数据集）。

00

总览

本文不仅收集了各式各样的数据集，和评估工具，并对其进行了按照不同标准的分类，供大家快速选择。

1. 评估工具

2. 主题

1. 里程计

2. 建图

3. 多样场景

4. 定位

5. 感知

3. 特点

1. 大规模

2. 长时间

3. 复杂地图

4. 极端情况

4. 载体平台

1. 汽车

2. 移动机器人

3. 无人机

4. 水下自主航行器

5. 水面无人艇

6. 手持设备

5. 环境

1. 城市

2. 室内

3. 地形

4. 水下

大家可以点击最后的原文链接查看本 GitHub Repo (PaoPaoRobot/awesome-slam-datasets) 来获得每个数据集的下载链接。

下图对每个数据集的来源机构，年份，环境，是否包含 GT pose，GT Map，IMU，GPS,语义标签，LIDAR，相机，事件相机，深度相机，雷达，声纳，多普勒速度记录，或其他相关信息进行了总结。

您可以在 repo 中查找到完整版表格或者根据此链接查看（需要科学上网）：

https://sites.google.com/view/awesome-slam-datasets/   

01

评估工具

• Trajectory Evaluation with Alignment

• Python package for the evaluation of odometry and SLAM

• SLAMBench2.0: SLAM performance evaluation framework

02

按主题分类

里程计

• TUM-Visual-Inertial

• Visual-Inertial Canoe Dataset

• Multi Vehicle Stereo Event Camera Dataset

• Zurich Urban Micro Aerial Vehicle Dataset

• EuRoC MAV Dataset

• TUM Monocular Cameras Dataset

• Event-Camera Dataset and Simulator

• TUM Omnidirectional Cameras Dataset

• ICL-NUIM RGBD Dataset

• TUM RGB-D SLAM Dataset and Benchmark

• Google Cartographer

• ADVIO Dataset

• Deep Inertial Odometry Dataset

• Aqualoc Underwater Dataset

• Rosario Agricultural Dataset

• Stereo Plenoptic Odometry Dataset

建图

• Collaborative SLAM Dataset (CSD)

• Complex Urban

• Multi-modal Panoramic 3D Outdoor Dataset (MPO)

• Underwater Caves SONAR and Vision Dataset

• Chilean Underground Mine Dataset

• Oxford Robotcar Dataset

• University of Michigan North Campus Long-Term (NCLT) Vision and LIDAR Dataset

• Málaga Stereo and Laser Urban Data Set

• KITTI Vision Benchmark Suite

• Challenging data sets for point cloud registration algorithms

• ACFR Marine Robotics Dataset

• Ford Campus Vision and Lidar Dataset

• InteriorNet

多样场景

• Visual-Inertial Canoe Dataset
  

• Symphony Lake Dataset

• Alderley Day/Night Dataset

• St Lucia Multiple Times of Day

• New College Vision and Laser Data Set

• FABMAP Dataset

定位

• Dataset for metric-level localization
  

• Cambridge Landmark Dataset

• KITTI Vision Benchmark Suite

• Microsoft 7 scenes

• San Francisco Landmark Dataset

感知

• KAIST Day/Night Dataset

• Robot @ Home Dataset

• SceneNet RBG-D Dataset

• Sugar Beets 2016, Agricultural Robot Dataset

• CityScapes Dataset

• KITTI Vision Benchmark Suite

• Multi-Sensor Perception (Marulan) Dataset

• InteriorNet

03

按特点分类

大规模

• Complex Urban
  

• Multi Vehicle Stereo Event Camera Dataset

• Multi-modal Panoramic 3D Outdoor Dataset (MPO)

• CityScapes Dataset

• Solar-powered UAV Sensing and Mapping Dataset

• Oxford Robotcar Dataset

• CCSAD (Stereo Urban) Dattaset

• Málaga Stereo and Laser Urban Data Set

• KITTI Vision Benchmark Suite

• Kagaru Airborne Stereo Dataset Dataset

• ACFR Marine Robotics Dataset

长时间

• KAIST Day/Night Dataset
  

• Visual-Inertial Canoe Dataset

• Symphony Lake Dataset

• Oxford Robotcar Dataset

• University of Michigan North Campus Long-Term (NCLT) Vision and LIDAR Dataset

• Alderley Day/Night Dataset

• St Lucia Multiple Times of Day

复杂地图

• Complex Urban
  

• Multi Vehicle Stereo Event Camera Dataset

• Multi-modal Panoramic 3D Outdoor Dataset (MPO)

• Málaga Stereo and Laser Urban Data Set

• KITTI Vision Benchmark Suite

• Challenging data sets for point cloud registration - algorithms

极端情况

• Underwater Caves SONAR and Vision Dataset: Underwater Environment
  

• Chilean Underground Mine Dataset: Underground Environment

• CityScapes Dataset: Foggy Scene

• EuRoC MAV Dataset: Fast motion

• Multi-Sensor Perception (Marulan) Dataset : Smoky, dust, and Rain condition

04

按载体平台分类

汽车

• Complex Urban Dataset

• Multi Vehicle Stereo Event Camera Dataset

• KAIST Day/Night Dataset

• Multi-modal Panoramic 3D Outdoor Dataset (MPO)

• Oxford Robotcar Dataset

• CityScapes Dataset

• CCSAD (Stereo Urban) Dattaset

• Málaga Stereo and Laser Urban Data Set

• KITTI Vision Benchmark Suite

• Day and Night with Lateral Pose Change Dataset

• Alderley Day/Night Dataset

• Annotated-laser Dataset (Link Broken)

• San Francisco Landmark Dataset

• Ford Campus Vision and Lidar Dataset

• St Lucia Stereo Vehicular Dataset

• St Lucia Multiple Times of Day

• MIT DARPA Urban Challenge Dataset

• FABMAP Dataset

移动机器人

• Rosario Dataset

• Sugar Beets 2016, Agricultural Robot Dataset

• Chilean Underground Mine Dataset

• Katwijk Beach Planetary Rover Dataset

• Robot @ Home Dataset

• University of Michigan North Campus Long-Term (NCLT) Vision and LIDAR Dataset

• Rawseeds In/Outdoor Dataset

• Canadian Planetary Emulation Terrain 3D Mapping Dataset

• Devon Island Rover Navigation Dataset

• Multi-Robot Cooperative Localization and Mapping Dataset

• Multi-Sensor Perception (Marulan) Dataset

• TUM RGB-D SLAM Dataset and Benchmark

• New College Vision and Laser Data Set

  
  

无人机

• Zurich Urban Micro Aerial Vehicle Dataset

• Event-Camera Dataset and Simulator

• Solar-powered UAV Sensing and Mapping Dataset

• EuRoC MAV Dataset

• Kagaru Airborne Stereo Dataset Dataset

水下自主航行器

• Aqualoc Underwater Dataset
  

• Underwater Caves SONAR and Vision Dataset

• ACFR Marine Robotics Dataset

水面无人艇

• Visual-Inertial Canoe Dataset
  

• Symphony Lake Dataset

手持设备

• Collaborative SLAM Dataset (CSD)

• SceneNet RBG-D Dataset

• Event-Camera Dataset and Simulator

• Comprehensive RGB-D Benchmark (CoRBS)

• Augmented ICL-NUIM Reconstruction Dataset

• ICL-NUIM RGBD Dataset

• Challenging data sets for point cloud registration algorithms

• Cosy Localization Database (COLD)

• ADVIO Dataset

• Deep Inertial Odometry Dataset

• InteriorNet

• Stereo Plenoptic Dataset

  
  

05

按环境分类

城市

• ADVIO Dataset
  

• Stereo Plenoptic Dataset

• KAIST Day/Night Dataset

• TUM-Visual-Inertial

• Complex Urban

• Multi Vehicle Stereo Event Camera Dataset

• Zurich Urban Micro Aerial Vehicle Dataset

• TUM Monocular Cameras Dataset

• CityScapes Dataset

• Oxford Robotcar Dataset

• University of Michigan North Campus Long-Term (NCLT) Vision and LIDAR Dataset

• Event-Camera Dataset and Simulator

• CCSAD (Stereo Urban) Dattaset

• TUM Omnidirectional Cameras Dataset

• Cambridge Landmark Dataset

• Málaga Stereo and Laser Urban Data Set

• KITTI Vision Benchmark Suite

• Alderley Day/Night Dataset

• Challenging data sets for point cloud registration algorithms

• Multi-Robot Cooperative Localization and Mapping Dataset

• Ford Campus Vision and Lidar Dataset

• San Francisco Landmark Dataset

• Annotated-laser Dataset (Link Broken)

• MIT DARPA Urban Challenge Dataset

• St Lucia Stereo Vehicular Dataset

• St Lucia Multiple Times of Day

• New College Vision and Laser Data Set

• Rawseeds In/Outdoor Dataset

• FABMAP Dataset

  
  

室内

• Collaborative SLAM Dataset (CSD)
  

• InteriorNet

• TUM-Visual-Inertial

• Multi-modal Panoramic 3D Outdoor Dataset (MPO)

• Robot @ Home Dataset

• SceneNet RBG-D Dataset

• EuRoC MAV Dataset

• TUM Monocular Cameras Dataset

• Comprehensive RGB-D Benchmark (CoRBS)

• TUM Omnidirectional Cameras Dataset

• Augmented ICL-NUIM Reconstruction Dataset

• ICL-NUIM RGBD Dataset

• Microsoft 7 scenes

• TUM RGB-D SLAM Dataset and Benchmark

• Cosy Localization Database (COLD)

• Rawseeds In/Outdoor Dataset

• Google Cartographer

  
  

地形

• Rosario Agricultural Dataset
  

• Visual-Inertial Canoe Dataset

• Chilean Underground Mine Dataset

• Symphony Lake Dataset

• Sugar Beets 2016, Agricultural Robot Dataset

• Katwijk Beach Planetary Rover Dataset

• Solar-powered UAV Sensing and Mapping Dataset

• Event-Camera Dataset and Simulator

• Canadian Planetary Emulation Terrain 3D Mapping Dataset

• Challenging data sets for point cloud registration - algorithms

• Kagaru Airborne Stereo Dataset Dataset

• Devon Island Rover Navigation Dataset

• Multi-Sensor Perception (Marulan) Dataset

  
  

水下

• Aqualoc Underwater Dataset
  

• Underwater Caves SONAR and Vision Dataset

• ACFR Marine Robotics Dataset

  
  

06

more

如果有更多相关数据集，可以向 Younggun Cho 的GitHub repo 提交 PR 或者 issue。

Repo 作者：Younggun Cho

翻译：Yvon Shong

欢迎来到泡泡论坛，这里有大牛为你解答关于SLAM的任何疑惑。

有想问的问题，或者想刷帖回答问题，泡泡论坛欢迎你！

泡泡网站：www.paopaorobot.org

泡泡论坛：http://paopaorobot.org/bbs/

泡泡机器人SLAM的原创内容均由泡泡机器人的成员花费大量心血制作而成，希望大家珍惜我们的劳动成果，转载请务必注明出自【泡泡机器人SLAM】微信公众号，否则侵权必究！同时，我们也欢迎各位转载到自己的朋友圈，让更多的人能进入到SLAM这个领域中，让我们共同为推进中国的SLAM事业而努力！

商业合作及转载请联系liufuqiang_robot@hotmail.com