自動駕駛研發需要蒐集大量數據，特別是採用深度學習訓練模型。雖然我們希望能夠解決在小數據或者少量數據的機器學習問題，但目前尚不能做到。這裡列出一些網上開源的數據集，供自動駕駛研發的同行，特別是創業公司，做參考。

標註數據是勞動密集行業，標註工具需要開發，也有些開源工具可以利用來二次開發，仍然有不少公司提供這種業務，比如矽谷灣區有一家MIT出來的Alex Wang創立的公司Scale.API，不久前還拿到了幾千萬美金的融資，它支持Nutonomy (purchased byDelphi-Aptiv)提供了開源數據NuScene。

1 Berkeley deep drive BDD 100k

http://bdd-data.berkeley.edu/.

2 Baidu Apolloscope Dataset

http://apolloscape.auto/

3 Comma.ai Dataset

https://archive.org/details/comma-dataset.

4 Oxford』s robotic car Dataset

http://robotcar-dataset.robots.ox.ac.uk/.

5 CityScapesDataSet

6 KITTI Dataset

The KITTI Vision Benchmark Suite

7 Ford Campus Vision and Lidar dataset

http://robots.engin.umich.edu/SoftwareData/Ford.

8 CamVid

?http://mi.eng.cam.ac.uk/research/projects/VideoRec/CamVid/

9 TuSimple Dataset

?http://benchmark.tusimple.ai/#/

10 CMU Visual Localization Dataset

?http://3dvis.ri.cmu.edu/data-sets/localization/.

11 NuScenes (Nutonomy-Aptiv dataset)

?https://www.nuscenes.org/

12 KUL Belgium Traffic Sign Dataset

http://www.vision.ee.ethz.ch/~timofter/traffic_signs/

13 MIT AGE Lab Dataset

http://lexfridman.com/carsync.

14 LISA: Laboratory for Intelligent & Safe Automobiles, UC San Diego Datasets

Laboratory for Intelligent and Safe Automobiles

15 Udacity challenge Datasets

?https://www.udacity.com/course/self-driving-car-engineer-nanodegree--nd013.

?https://github.com/udacity/self-driving-car/tree/master/datasets.

?http://academictorrents.com/userdetails.php?id=5125.

?https://github.com/rwightman/udacity-driving-reader.

16 VelodyneSLAM Dataset

?http://www.mrt.kit.edu/z/publ/download/velodyneslam/dataset.html

17 The UAH-DriveSet

?http://www.robesafe.uah.es/personal/eduardo.romera/uah-driveset/

18 Berkeley DeepDrive (BDD) Driving Model

?https://github.com/gy20073/bdd_driving_model

?http://bdd-data.berkeley.edu/wad-2018.html

19 Berkeley DeepDrive (BDD) Driving Model

?https://github.com/gy20073/bdd_driving_model

?http://bdd-data.berkeley.edu/wad-2018.html

20 MIT-AVT: Autonomous Vehicle Technology

?https://hcai.mit.edu/avt/;

21 Virtual KITTI Dataset

?http://www.europe.naverlabs.com/Research/Computer-Vision/Proxy-Virtual-Worlds.

22 Synthia Dataset

http://synthia-dataset.net/

23 FAB-MAP: Probabilistic Localization And Mapping In The Space Of Appearance

Oxford Robotcar Dataset

24 A High-rate, Heterogeneous Data Set From The DARPA Urban Challenge

?http://grandchallenge.mit.edu/wiki/index.php?title=publicdata

25 The Marulan Data Sets: Multi-sensor Perception In Natural Environment With Challenging Conditions

?http://sdi.acfr.usyd.edu.au/;

26 Unaided Stereo Vision Based Pose Estimation with UQ St Lucia Dataset

?http://asrl.utias.utoronto.ca/~mdw/uqstluciadataset.html

27 A Ma』LagaCollection Of Outdoor Robotic Datasets With Centimeter-accuracy Ground Truth

?https://www.mrpt.org/malaga_dataset_2009

28 The Ma』Laga Urban Dataset: High-rate Stereo And Lidars In A Realistic Urban Scenario

?http://www.mrpt.org/malagaurbandataset

29 Complex Urban Lidar Data Set from KAIST

?http://irap.kaist.ac.kr/dataset/

30 KAIST Multi-spectral Day/Night Data Set For Autonomous And Assisted Driving

?https://sites.google.com/view/multispectral/ (coming soon)

31 L-CAS 3D Point Cloud People Dataset

L-CAS 3D Point Cloud People Dataset?

lcas.lincoln.ac.uk

具體不解釋了。

推薦閱讀：