Open datasets

Discover the best
free datasets
for machine learning

Browse all the open and public datasets hosted on and off Segments.ai. You can easily clone and adapt them to your own needs.

Did you create a public dataset that you want to share?
Reach out to us.

multi sensor data fusion labeling platform for machine learning engineers
Filters
Dataset Type
Dataset Task
Out- or indoor
Search
  • BDD 100K

    Image

    A diverse driving dataset for heterogeneous multitask learning

    100M

    frames

    40

    classes

    100%

    labeled

  • nuPlan

    Multi-sensor

    nuPlan is a large-scale planning benchmark for autonomous driving.

    38.8M

    frames

    7

    classes

    100%

    labeled

  • Waymo Motion

    Point cloud > Bounding box

    High-resolution sensor data collected by autonomous vehicles operated by the Waymo Driver in a wide variety of conditions.

    20M

    frames

    3

    classes

    100%

    labeled

  • Argoverse 2 Sensor

    Multi-sensor

    A collection of open-source autonomous driving data from six U.S. cities.

    6M

    frames

    30

    classes

    100%

    labeled

  • nuScenes

    Multi-sensor

    nuScenes is a dataset for autonomous driving. It includes street scenes from Singapore and Boston, U.S.

    1.8M

    frames

    32

    classes

    100%

    labeled

  • A2D2

    Multi-sensor

    A2D2 stands for Audi Autonomous Driving Dataset (A2D2). The data was captured in 3 German cities.

    400K

    frames

    38

    classes

    10%

    labeled

  • Waymo Perception

    Multi-sensor

    Diverse autonomous driving dataset with scenes captured in 6 U.S. areas with different environments and weather conditions.

    390K

    frames

    28

    classes

    100%

    labeled

  • Zenseact

    Multi-sensor

    A large multi-modal autonomous driving dataset, created by researchers at Zenseact.

    100K

    frames

    30

    classes

    100%

    labeled

  • JackRabbot Dataset and Benchmark

    Multi-sensort

    Indoor and outdoor data from a human perspective.

    60K

    frames

    1

    classes

    100%

    labeled

  • PandaSet

    Multi-sensor

    High-quality dataset by lidar producer Hesai. Its 100+ scenes are selected from two routes in Silicon Valley.

    48K

    frames

    37

    classes

    100%

    labeled

  • ApolloScape

    Point cloud > Bounding box

    Autonomous driving dataset created by Baidu research, under various lighting conditions and traffic densities in Beijing.

    31.8K

    frames

    6

    classes

    20%

    labeled

  • Leddar PixSet

    Multi-sensor

    The PixSet dataset contains 97 sequences for a total of roughly 29k frames using the AV sensor suite.

    29K

    frames

    22

    classes

    100%

    labeled

  • DENSE Seeing Through Fog

    Multi-sensor > Bounding box

    Different weather conditions like fog, snow, and rain; captured in northern Europe.

    12K

    frames

    4

    classes

    100%

    labeled

  • KITTI

    Multi-sensor

    KITTI is a dataset of lidar sequences of street scenes in Karlsruhe, Germany.

    7.5K

    frames

    8

    classes

    100%

    labeled