Data Structures¶

This section provides an overview of all data structures in the MIMIC Data API.

Core Classes¶

Scene¶

The top-level container for all scene data.

Attribute	Description
`ego_vehicle`	The recording/autonomous vehicle with sensors and trajectory
`actors`	List of all other objects in the scene

See Scene Module for full documentation.

EgoVehicle¶

Represents the ego (recording) vehicle.

Attribute	Description
`trajectory`	Time-series of ego poses in world coordinates
`sensors`	Collection of mounted sensors (cameras, lidars, radars)

Sensors¶

Container for all sensor data.

Attribute	Description
`cameras`	List of camera sensors (required)
`lidars`	List of LiDAR sensors (optional)
`radars`	List of radar sensors (optional)

Trajectory¶

WorldTrajectory¶

Time-indexed sequence of poses in world coordinates.

Attribute	Description
`timestamps`	1D array of N timestamps in seconds
`object2world`	(N, 4, 4) array of transformation matrices
`linear_velocity`	Optional (N, 3) array of velocities in m/s
`angular_velocity`	Optional (N, 3) array of angular velocities in rad/s

See Trajectory Module for full documentation.

Camera Structures¶

Camera¶

Complete camera sensor representation.

Attribute	Description
`id`	Unique identifier
`name`	Human-readable name (e.g., “front_camera”)
`camera_model`	Projection model type (pinhole, opencv, fisheye, etc.)
`intrinsics`	Internal parameters (focal length, principal point)
`distortion_parameters`	Lens distortion coefficients
`extrinsics`	Sensor-to-vehicle transformation matrix
`trajectory`	Camera poses over time
`images_data`	List of captured images

CameraIntrinsics¶

Internal camera calibration parameters.

Attribute	Description
`fx`, `fy`	Focal lengths in pixels
`cx`, `cy`	Principal point coordinates
`skew`	Skew coefficient (usually 0)
`width`, `height`	Image dimensions in pixels

DistortionParameters¶

Lens distortion coefficients supporting multiple camera models.

Attribute	Description
`k1`, `k2`, `k3`, `k4`	Radial distortion coefficients
`p1`, `p2`	Tangential distortion coefficients
`k5`, `k6`	Extended radial coefficients (optional)
`s1`, `s2`, `s3`	Thin prism coefficients (optional)

ImageData¶

Metadata for a single captured image.

Attribute	Description
`filepath`	Path to image file (PNG or JPEG)
`timestamp`	Capture time in seconds
`undistorded`	Whether distortion has been removed
`rolling_shutter`	Rolling shutter parameters

See Camera Module for full documentation.

LiDAR Structures¶

Lidar¶

Complete LiDAR sensor representation.

Attribute	Description
`id`	Unique identifier
`name`	Human-readable name (e.g., “top_lidar”)
`intrinsics`	Sensor parameters (FOV, resolution, range)
`extrinsics`	Sensor-to-vehicle transformation
`trajectory`	LiDAR poses over time
`lidar_pcds_data`	List of point cloud captures
`min_cut_off_range`	Ego vehicle filter threshold

LidarIntrinsics¶

Internal LiDAR parameters.

Attribute	Description
`lidar_type`	“rotating” or “solid-state”
`rotation_type`	“clockwise” or “counter-clockwise”
`vertical_fov`	Vertical field of view in radians
`horizontal_fov`	Horizontal field of view in radians
`horizontal_resolution`	Radians per point (angular step between azimuth samples), e.g. `2*pi/num_points_per_rotation`. Not points per radian.
`vertical_resolution`	Vertical resolution specification
`max_range`, `min_range`	Detection range limits in meters

VerticalResolution¶

Vertical angular resolution specification.

Attribute	Description
`vertical_resolution`	Uniform resolution in radians per point (angular step), not points/radian
`elevation_channel_mapping`	Per-channel angle mapping (alternative)

Note: Exactly one of these must be provided.

LidarPointCloudData¶

Metadata for a single LiDAR scan.

Attribute	Description
`filepath`	Path to PLY file
`timestamp`	Mid-scan timestamp in seconds
`egomotion_compensated`	Whether motion compensation applied
`has_missing_points`	Whether invalid points exist

See LiDAR Module for full documentation.

Radar Structures¶

Radar¶

Complete radar sensor representation.

Attribute	Description
`id`	Unique identifier
`name`	Human-readable name
`intrinsics`	Sensor parameters
`extrinsics`	Sensor-to-vehicle transformation
`trajectory`	Radar poses over time
`radar_pcds_data`	List of radar detections

RadarIntrinsics¶

Internal radar parameters.

Attribute	Description
`range`	Maximum detection range in meters
`azimuth_fov`	Horizontal FOV in radians
`elevation_fov`	Vertical FOV in radians
`azimuth_resolution`	Angle-dependent horizontal resolution
`elevation_resolution`	Vertical resolution (imaging radars)
`min_speed_resolution`	Minimum Doppler velocity (m/s)
`max_speed_resolution`	Maximum Doppler velocity (m/s)

See Radar Module for full documentation.

Actor Structures¶

Actor¶

Representation of scene objects (vehicles, pedestrians, etc.).

Attribute	Description
`id`	Unique identifier
`name`	Optional human-readable name
`dimensions`	Physical bounding box size
`boxtype`	Category: “vehicle”, “pedestrian”, “bicycle”, “sign”, “unknown”
`original_boxtype`	Original dataset label (optional)
`dynamic`	True if moving, False if static
`trajectory`	Actor poses over time
`description`	Optional free-text description

ActorDimensions¶

Physical dimensions of an actor.

Attribute	Description
`length`	Length in meters (forward axis)
`width`	Width in meters (lateral axis)
`height`	Height in meters (vertical axis)

See Actor Module for full documentation.

Type Aliases¶

SensorExtrinsic¶

A 4x4 NumPy array representing the sensor-to-vehicle transformation matrix:

SensorExtrinsic = NDArray[Shape["4 x, 4 y"], float]

Camera Models¶

Supported camera projection models:

CameraModels = Literal[
    "simple_pinhole",
    "pinhole",
    "simple_radial",
    "radial",
    "opencv",
    "opencv_fisheye",
    "equirectangular",
    "omnidirectionalstereo_l",
    "omnidirectionalstereo_r",
    "vr180_l",
    "vr180_r",
    "orthophoto",
    "fisheye624",
]

These models are aligned with COLMAP conventions.