Towards Self-driving Vehicles

Edit this page

Datasets for Self-Driving Vehicles

  • Udacity - Udacity driving datasets released for Udacity Challenges. Contains ROSBAG training data. (~80 GB).
  • Comma.ai - 7 and a quarter hours of largely highway driving. Consists of 10 videos clips of variable size recorded at 20 Hz with a camera mounted on the windshield of an Acura ILX 2016. In parallel to the videos, also recorded some measurements such as car’s speed, acceleration, steering angle, GPS coordinates, gyroscope angles. These measurements are transformed into a uniform 100 Hz time base.
  • Oxford’s Robotic Car - over 100 repetitions of a consistent route through Oxford, UK, captured over a period of over a year. The dataset captures many different combinations of weather, traffic and pedestrians, along with longer term changes such as construction and roadworks.
  • KITTI Vision Benchmark Suite - 6 hours of traffic scenarios at 10-100 Hz using a variety of sensor modalities such as highresolution color and grayscale stereo cameras, a Velodyne 3D laser scanner and a high-precision GPS/IMU inertial navigation system.
  • University of Michigan North Campus Long-Term Vision and LIDAR Dataset - consists of omnidirectional imagery, 3D lidar, planar lidar, GPS, and proprioceptive sensors for odometry collected using a Segway robot.
  • University of Michigan Ford Campus Vision and Lidar Data Set - dataset collected by an autonomous ground vehicle testbed, based upon a modified Ford F-250 pickup truck. The vehicle is outfitted with a professional (Applanix POS LV) and consumer (Xsens MTI-G) Inertial Measuring Unit (IMU), a Velodyne 3D-lidar scanner, two push-broom forward looking Riegl lidars, and a Point Grey Ladybug3 omnidirectional camera system.
  • DIPLECS Autonomous Driving Datasets (2015) - dataset was recorded by placing a HD camera in a car driving around the Surrey countryside. The dataset contains about 30 minutes of driving. The video is 1920x1080 in colour, encoded using H.264 codec. Steering is estimated by tracking markers on the steering wheel. The car’s speed is estimated from OCR the car’s speedometer (but the accuracy of the method is not guaranteed).
  • Velodyne SLAM Dataset from Karlsruhe Institute of Technology - two challenging datasets recorded with the Velodyne HDL64E-S2 scanner in the city of Karlsruhe, Germany.
  • SYNTHetic collection of Imagery and Annotations (SYNTHIA) - consists of a collection of photo-realistic frames rendered from a virtual city and comes with precise pixel-level semantic annotations for 13 classes: misc, sky, building, road, sidewalk, fence, vegetation, pole, car, sign, pedestrian, cyclist, lanemarking.
  • Cityscape Dataset - focuses on semantic understanding of urban street scenes. large-scale dataset that contains a diverse set of stereo video sequences recorded in street scenes from 50 different cities, with high quality pixel-level annotations of 5 000 frames in addition to a larger set of 20 000 weakly annotated frames. The dataset is thus an order of magnitude larger than similar previous attempts. Details on annotated classes and examples of our annotations are available.
  • CSSAD Dataset - Several real-world stereo datasets exist for the development and testing of algorithms in the fields of perception and navigation of autonomous vehicles. However, none of them was recorded in developing countries and therefore they lack the particular characteristics that can be found in their streets and roads, like abundant potholes, speed bumpers and peculiar flows of pedestrians. This stereo dataset was recorded from a moving vehicle and contains high resolution stereo images which are complemented with orientation and acceleration data obtained from an IMU, GPS data, and data from the car computer.
  • Daimler Urban Segmetation Dataset - consists of video sequences recorded in urban traffic. The dataset consists of 5000 rectified stereo image pairs with a resolution of 1024x440. 500 frames (every 10th frame of the sequence) come with pixel-level semantic class annotations into 5 classes: ground, building, vehicle, pedestrian, sky. Dense disparity maps are provided as a reference, however these are not manually annotated but computed using semi-global matching (sgm).
  • Self Racing Cars - XSens/Fairchild Dataset - The files include measurements from the Fairchild FIS1100 6 Degree of Freedom (DoF) IMU, the Fairchild FMT-1030 AHRS, the Xsens MTi-3 AHRS, and the Xsens MTi-G-710 GNSS/INS. The files from the event can all be read in the MT Manager software, available as part of the MT Software Suite, available here.
  • MIT AGE Lab - a small sample of the 1,000+ hours of multi-sensor driving datasets collected at AgeLab.
  • Yet Another Computer Vision Index To Datasets (YACVID) - a list of frequently used computer vision datasets.
  • KUL Belgium Traffic Sign Dataset - a large dataset with 10000+ traffic sign annotations, thousands of physically distinct traffic signs. 4 video sequences recorded with 8 high resolution cameras mounted on a van, summing more than 3 hours, with traffic sign annotations, camera calibrations and poses. About 16000 background images. The material is captured in Belgium, in urban environments from Flanders region, by GeoAutomation.
  • LISA: Laboratory for Intelligent & Safe Automobiles, UC San Diego Datasets - traffic sign, vehicles detection, traffic lights, trajectory patterns.
  • Multisensory Omni-directional Long-term Place Recognition (MOLP) dataset for autonomous driving It was recorded using omni-directional stereo cameras during one year in Colorado, USA. paper
  • Lane Instance Segmentation in Urban Environments Semi-automated method for labelling lane instances. 24,000 image set available. paper
  • Foggy Zurich Dataset Curriculum Model Adaptation with Synthetic and Real Data for Semantic Dense Foggy Scene Understanding. 3.8k High Quality Foggy images in and around Zurich. paper
  • SullyChen AutoPilot Dataset Dataset collected by SullyChen in and around California.
  • Waymo Training and Validation Data One terabyte of data with 3D and 2D labels.
  • Intel’s dataset for AD conditions in India A dataset for Autonomous Driving conditions in India with segmented annotations (10k). (by Intel & IIIT Hyderabad).
  • nuScenes Dataset A large dataset with 1,400,000 images and 390,000 lidar sweeps from Boston and Singapore. Provides manually generated 3D bounding boxes for 23 object classes.
  • German Traffic Sign Dataset A large dataset of German traffic sign recogniton data (GTSRB) with more than 40 classes in 50k images and detection data (GTSDB) with 900 image annotations.
  • Swedish Traffic Sign Dataset A dataset with traffic signs recorded on 350 km of Swedish roads, consisting of 20k+ images with 20% of annotations.

Note:

  • This page was originally taken from Awesome Autonomous Vehicles.
  • TODO: Divide into different tasks, such as lane line segmentation, traffic sign detection…