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
Filtered (0)
Filters
Dataset Type
Dataset Task
Out- or indoor
Search
Show (0)
Cancel

  • BDD 100K

    Image

    A diverse driving dataset for heterogeneous multitask learning

    100M

    frames

    40

    classes

    100%

    labeled

  • Multiple camera feeds from nuPlan showcase various views of a city's streets and intersections with identified objects and vehicles outlined.

    nuPlan

    Multi-sensor

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

    38.8M

    frames

    7

    classes

    100%

    labeled

  • Aerial view of LiDAR technology identifying vehicles and objects on a street in motion, highlighted with red boxes.

    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

  • A collage of LiDAR data from advanced sensor technology, street scenes with annotated vehicles, and two bar graphs comparing Argoverse 1 and 2 datasets.

    Argoverse 2 Sensor

    Multi-sensor

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

    6M

    frames

    30

    classes

    100%

    labeled

  • Six top-view renderings of various rooms, showcasing different furniture arrangements in vibrant colors, created using advanced scanning technology from ScanNet.

    ScanNet

    Point cloud > Segmentation

    Indoor scenes containing 2.5 million views in 1513 scans

    2.5M

    frames

    40

    classes

    90%

    labeled

  • A 3D point cloud image with various colored data points representing an urban environment, overlaid on street view images, useful for autonomous driving and derived from the KITTI dataset.

    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

  • Several cars waiting at a signalized intersection with a yellow road sign in the background, surrounded by trees, reminiscent of scenes used in A2D2 datasets.

    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

  • 3D point cloud image of an outdoor space featuring trees, steps, and pathways, captured using LiDAR technology by Waymo for enhanced perception.

    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

  • A digitally distorted urban street scene in Toronto with buildings, cars, and traffic lights visible amidst 3D artifact effects.

    Toronto-3D

    Point cloud

    Detailed dataset of 1km of road in Toronto, Canada.

    249K

    frames

    8

    classes

    100%

    labeled

  • DrivingStereo

    Multi-sensor

    A Large-Scale Dataset for Stereo Matching in Autonomous Driving Scenarios

    180K

    frames

    classes

    0%

    labeled

  • Street view with multiple dynamic objects annotated in bounding boxes, showing cars, pedestrians, and traffic lights.

    Zenseact

    Multi-sensor

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

    100K

    frames

    30

    classes

    100%

    labeled

  • An image showing a 360-degree camera view with 2D and 3D bounding boxes and point cloud overlay detecting multiple people, utilizing data from the JackRabbot Benchmark.

    JackRabbot Dataset and Benchmark

    Multi-sensort

    Indoor and outdoor data from a human perspective.

    60K

    frames

    1

    classes

    100%

    labeled

  • Black and white LIDAR scan of an outdoor area, displaying detailed point clouds and yellow highlighted regions. Captured using PandaSet for enhanced precision and clarity.

    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's digital radar display features blue circular waves at the center, flanked by numbers in small black boxes and green data points.

    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

  • Top: 3D point cloud data with vehicles and buildings from PixSet. Bottom: Street view with overlaid object detection boxes.

    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

  • A busy urban intersection is alive with cars navigating through, billboards towering above, and large store signs on buildings like McDonald's, Koton, and Bershka. This vibrant scene offers dynamic vistas that could easily be captured through Mapillary's street-level imagery, showcasing the hustle of city life.

    Mapillary Vistas

    Image > Segmentation

    street-level imagery dataset with pixel‑accurate and instance‑specific human annotations.

    2.5K

    frames

    124

    classes

    100%

    labeled

  • A vibrant 3D LIDAR point cloud image of an urban street with vehicles and buildings, featuring a circular roundabout, is enriched with SemanticKITTI data for enhanced SEO.

    SemanticKITTI

    Point cloud

    SemanticKITTI is a dataset of LiDAR sequences for semantic scene understanding.

    23.2K

    frames

    28

    classes

    100%

    labeled

  • Top row: horse parade, two men playing video games, people by water. Bottom row: COCO outlined versions of top images.

    COCO

    Image > Segmentation

    200K labeled images of objects such as different kinds of animals, appliances, food, and much more.

    20K

    frames

    80

    classes

    100%

    labeled

  • Rooms

    Image > Segmentation

    A dataset of rooms

    12.9K

    frames

    58

    classes

    0%

    labeled

  • Composite image with four views (stereo, gated, thermal, lidar) of a car and traffic lights on a foggy road, showcasing the technology's power in seeing through fog.

    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

  • saptadeb – Soy

    Image > Segmentation

    A dataset of soy fields

    11.2K

    frames

    1

    classes

    2%

    labeled

  • A highway view from inside a car, with identified cars and road areas outlined and highlighted in different colors, using comma10k technology.

    comma10k

    Image > Segmentation

    10,000 PNGs of real driving captured from the comma fleet, semantically labeled by the public.

    10K

    frames

    5

    classes

    93%

    labeled

  • Street view with bounding boxes and lidar data overlays from a self-driving car project, including KITTI and CVLIBS logos.

    KITTI

    Multi-sensor

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

    7.5K

    frames

    8

    classes

    100%

    labeled

  • saptadeb – Corn

    Image > Segmentation

    A dataset of corn fields

    6.4K

    frames

    1

    classes

    32%

    labeled

  • Cityscapes

    Image > Segmentation

    The Cityscapes Dataset focuses on semantic understanding of urban street scenes.

    5K

    frames

    classes

    100%

    labeled

  • Hallway Gyeonggibuk Science High

    Image > Segmentation

    Hallway on the 3rd floor of Gyeonggibuk Science High School.

    3K

    frames

    1

    classes

    60%

    labeled

  • MassMIND: Massachusetts Maritime INfrared Dataset image segmentation - A silhouetted person fishing.

    MassMIND: Massachusetts Maritime INfrared Dataset

    Segmentation

    3K

    frames

    7

    classes

    100%

    labeled

  • Abstract colorful image with a central pathway, trees, and various splashes of green, yellow, red, and blue, evoking a sense of winter's vibrant contrast.

    Winter Adverse Driving

    Point cloud

    20 scenes for autonomous driving collected in severe winter weather in Michigan, U.S.

    2K

    frames

    22

    classes

    27%

    labeled

  • Intersection of a city street with tall buildings in the background, segmented by colors orange, green, and blue, resembling a scene from SemanticKITTI.

    Panoptic – Visible

    Image > Segmentation

    A panoptic road segmentation

    2K

    frames

    133

    classes

    100%

    labeled

  • A suburban street in autumn with orange-tinted grass and trees, fallen leaves scattered on the ground, resembles scenes captured in the KITTI dataset used for Semantic Segmentation.

    Lawn mower

    Image > Segmentation

    Driveable area on lawns

    1.3K

    frames

    1

    classes

    100%

    labeled

  • A street with cars, adjacent bike lane, sidewalk, and separated areas, highlighted in various colors for segmentation using SemanticKITTI.

    Sidewalk Segmentation

    Image > Segmentation

    A dataset of sidewalk images gathered in Belgium in the summer of 2021

    1K

    frames

    34

    classes

    100%

    labeled

  • A car waits at a red light on a tree-lined urban street with buildings and traffic signs highlighted in various colors, an ideal scene for testing SemanticKITTI's capabilities.

    Manthan

    Image > Segmentation

    A dataset consisting of road scenes

    641

    frames

    4

    classes

    48%

    labeled

  • A red SUV parked in an empty outdoor lot, with a yellowish tint to the background, captured the scene reminiscent of data points from SemanticKITTI.

    Cars background removal

    Image > Segmentation

    A dataset of cars

    617

    frames

    3

    classes

    57%

    labeled

  • A colorful 3D LIDAR scan of a street scene, demonstrated with Ouster's OS 2 technology, showcasing cars and buildings rendered in vibrant red, green, blue, and purple points.

    Ouster OS 0 Demo

    Point cloud > Vector

    OS0 is an ultra wide lidar

    500

    frames

    2

    classes

    0%

    labeled

  • A colorful 3D LIDAR point cloud image of an outdoor circular area with trees and surrounding structures visible, captured during a demo using the Ouster OS 2 sensor.

    Ouster OS 1 Demo

    Point cloud > Vector

    OS1 is a mid-range lidar

    500

    frames

    2

    classes

    0%

    labeled

  • Colorful 3D point cloud map demo showcasing various structures and trees, overlaid with a neon-like, rainbow color gradient captured by the Ouster OS 2.

    Ouster OS 2 Demo

    Point cloud > Vector

    OS2 is a long-range lidar

    500

    frames

    2

    classes

    0%

    labeled

  • A hallway with doors on both sides. A fire extinguisher, a small container, and a box are placed on the floor, hinting at nearby classrooms dedicated to teaching and education.

    Class

    Image > Segmentation

    A dataset of a hallway with objects obstructing the way.

    451

    frames

    3

    classes

    100%

    labeled

  • A city street with buildings, trees, and a bus. The image is color-segmented into distinct areas, highlighting recent asphalt repair and road damage caused by potholes.

    Vienna Automotive Dataset

    Image > Segmentation

    Roadscene dataset in Vienna

    295

    frames

    29

    classes

    8%

    labeled

  • Split screen showcasing a road view from a car dashboard, with a SemanticSpray radar scan of the same scene below it.

    SemanticSpray

    Scenes in wet surface conditions captured by camera, LiDAR and radar.

    200

    frames

    10

    classes

    100%

    labeled

  • A road, undergoing pavement maintenance to address road damage, is surrounded by trees and a building on the left side. It leads into a forested area with mountains in the background.

    Lane lines

    Image > Segmentation

    Lane lines segmentation on a road

    189

    frames

    1

    classes

    100%

    labeled

  • Room with a table, boxes, standing fan, and ladder against a purple wall. The floor is brown, the table and boxes are yellow, and the walls look like they've seen more street repair than decor.

    Youbot Semantics

    Image > Segmentation

    Indoor segmentation from the viewpoint of the YouBot robot

    140

    frames

    13

    classes

    100%

    labeled

  • A dirt path with scattered potholes leading between tall, dense hedges in a farm or garden setting, extending into the distance on an overcast day.

    Blackberry2Right

    Image > Segmentation

    Segmentation of driveable space in a field of blackberries

    100

    frames

    1

    classes

    98%

    labeled

  • Aerial view of residential neighborhood with orange lines highlighting various pathways or roads, indicating areas designated for road maintenance.

    Road satelite images

    Image > Segmentation

    Road segmentation for satelite images

    97

    frames

    1

    classes

    100%

    labeled

  • A point cloud of a strawberry plant, thriving in a small black container on a wooden table.

    LAST-Straw

    Point Cloud > Segmentation

    A temporal dataset of strawberry plant point clouds at different growth and maturity stages.

    84

    frames

    9

    classes

    33%

    labeled

  • Pavement with road damage on a tree-lined street in daylight.

    Potholes

    Image > Segmentation

    A dataset of potholes

    81

    frames

    1

    classes

    100%

    labeled