Autonomous vehicle datasets

There are a lot of collections of datasets, but unfortunately the majority of these datasets are not truly FREE to use. They are only for research purposes.

Here we'll list all datasets that have an MIT-type license, aka they are allowed to be used for both research, hobby and commercial purposes.

Data is mostly considered one of the most valuable things for a robotic company, however I think even using smaller datasets as a baseline for further fine-tuning or comparing to without ever worrying about any license issues are worth pursuing. Truly open knowledge and data is the future.

Lidar

Dataset	License	Annotations	Year	Link

Vision

Dataset	License	Annotations	Link
Caltech Pedestrian detection	CC4.0	Pedestrian bbox	http://www.vision.caltech.edu/Image_Datasets/CaltechPedestrians/index.html
GM-ATCI Rear-view pedestrians	CC4.0/MIT [1]		https://sites.google.com/site/rearviewpeds1/
SullyChen AutoPilot Dataset	MIT	Steering angle	https://github.com/SullyChen/Autopilot-TensorFlow
COCO	CC4.0	Segmentation	http://cocodataset.org/
Self driving car data	CC0	None	https://www.kaggle.com/ajaysh/self-driving-car
CSAIL LabelMe dataset	CC0 (with attribution)	Segmentation	http://labelme2.csail.mit.edu/Release3.0/browserTools/php/publications.php
Udacity	MIT	Bounding boxes	https://github.com/udacity/self-driving-car/tree/master/annotations
Udacity (Roboflow cleaned)	MIT	Bounding boxes	https://public.roboflow.ai/object-detection/self-driving-car
NGSIM Vehicle Trajectories and Supporting Data	CC4.0	Vehicle trajectory data	https://data.transportation.gov/Automobiles/Next-Generation-Simulation-NGSIM-Vehicle-Trajector/8ect-6jqj
ADE20K	BSD 3.0	Semantic segmentation	http://groups.csail.mit.edu/vision/datasets/ADE20K/index.html

[1] Confirmed with GM per email. No responsibility whatsoever from GM. So something similar to MIT

Multi-sensor

Dataset	License	Annotations	Sensors	Link
Udacity	MIT	None	IMU/GPS/LIDAR	https://github.com/udacity/self-driving-car/tree/master/datasets
Velodyne SLAM KIT	CC4.0 [2]	None	LIDAR/Stereo camera	https://www.mrt.kit.edu/z/publ/download/velodyneslam/dataset.html
NCLT Dataset	Open Database License		Camera/IMU/GPS/LIDAR/Wheel odometry	http://robots.engin.umich.edu/nclt/
Brno Urban Dataset	MIT	None	Camera/Thermal/RTK/GPS/LIDAR/IMU	https://github.com/Robotics-BUT/Brno-Urban-Dataset
Audi Autonomous Driving Dataset (A2D2)	CC BY-ND 4.0	3D bounding boxes/Semantic Segmentation	Camera/LIDAR	https://www.a2d2.audi/
comma2k19	MIT	None	Camera/IMU/GPS	https://github.com/commaai/comma2k19
PedX	MIT	3D segmentation/2D labels	Camera/LIDAR	http://pedx.io/
Scale AI PandaSet	CC4.0	3D Bounding boxes/PointCloud annotations	Lidar/Camera	https://scale.com/resources/download/pandaset
[2] This data can be freely used with one restriction: In case it is used for scientific publication
you are required to cite the article "Velodyne SLAM" of Moosmann et al

Synthetic data

Dataset	License	Annotations	Year	Link
Semantic Segmentation for Self Driving Cars	CC0	semantic segmenation		https://www.kaggle.com/kumaresanmanickavelu/lyft-udacity-challenge

References

https://www.datasetlist.com/

About

Autonomous driving datasets that are free to use commercially (MIT)

MIT License