Stardust - Data structure (v1.0)

Download OpenAPI specification:Download

Welcome to Stardust Data, here is our import/export data structure documentation

[中文版本] [English Version]

PointCloud

PointCloud import data structure

Please give the data in the following format, which will facilitate us to adapt the data at the fastest speed

for single frame and continuous frames, .json The file is an independent parameter for each camera, and the configurable content refers to Camera parameters Partial

  -data_root
    -Camera1
        1620724790.jpg
        1620724791.jpg
        ...
     Camera1.json

    -Camera2
        1620724790.jpg
        1620724791.jpg
        ...
     Camera2.json

    -Camera3
        1620724790.jpg
        1620724791.jpg
        ...
     Camera3.json

    -Lidar
        1620724790.pcd
        1620724791.pcd
        ...

Camera parameters

Coordinate system description: The laser radar coordinate system is used, and the forward direction of the vehicle is taken as thexthe axis's positive direction's right-handed system, if the parameter coordinate system does not correspond to it, it needs to be transformed and given to

The external parameter adopts the transformation matrix or specific parameters of the camera relative to the laser radar in the lidar coordinate system. If it is the external parameter matrix of the radar relative to the camera coordinate system, it needs to be processed by inverse transformation and then given to speed up the data adaptation. process

for corresponding camera folder namingjsonfile, The content is The detailed parameters after decomposition

fov
number
required
object (CameraHeading)

Camera orientation, relative to the origin in the point cloud file (0, 0, 0) orientation, quaternion

required
object (CameraIntrinsicParameter)

Camera intrinsic parameters

matrix
Array of numbers[ items ]

Bev mapping matrix

maxz
number
minz
number
required
object (PointCloudCoordinate)

Camera position, relative to the origin in the point cloud file (0, 0, 0) s position

projection
Array of numbers[ items ]
required
object (CameraRadial)

Camera Radial Distortion Parameters

skew
required
number

camera distortion factor

required
object (CameraTangential)

Camera Tangential Distortion Parameters

type
required
string (CameraType)
Enum: "PinHole" "Fisheye" "OmniDirectional" "Bev"

Camera type

unProjection
Array of numbers[ items ]
{
  • "type": "PinHole",
  • "heading": {
    },
  • "position": {
    },
  • "intrinsic": {
    },
  • "skew": 0,
  • "radial": {
    },
  • "tangential": {
    },
  • "imageWidth": 1080,
  • "imageHeight": 960
}

Point Cloud

required
object (PointCloudSource)

Single frame point cloud data source

attachmentType
required
string

The single frame point cloud input type isPOINTCLOUD

Value: "POINTCLOUD"
extraInfoUrls
Array of strings

Additional reference information can be url link or text, needs to be aArray

object (PointCloudDataRecordMetadata)

Point cloud annotation may be used metadata information

{
  • "attachmentType": "POINTCLOUD",
  • "metadata": {
    },
  • "attachment": {
    }
}

PointCloud Sequence

required
Array of objects (PointCloudSource) [ items ]

point cloud sequence data source

attachmentType
required
string

The point cloud sequence input type isPOINTCLOUD_SEQUENCE

Value: "POINTCLOUD_SEQUENCE"
extraInfoUrls
Array of strings

Additional reference information can be url link or text, needs to be aArray

object (PointCloudDataRecordMetadata)

Point cloud annotation may be used metadata information

{
  • "attachmentType": "POINTCLOUD_SEQUENCE",
  • "metadata": {
    },
  • "attachment": [
    ]
}

Image

Image

required
string or ImageSourceEntity (object) or Array of ImageSourceEntity (objects) (ImageSource)

Single frame picture input data link

attachmentType
required
string

Single frame picture input type isIMAGE

Value: "IMAGE"
extraInfoUrls
Array of strings

Additional reference information can be url link or text, needs to be aArray

object (ImageDataRecordMetadata)

Image annotation may be used metadata information

{}

Image Sequence

required
Array of strings or ImageSourceEntity (object) or Array of ImageSourceEntity (objects) (ImageSource) [ items ]

Image sequence input data link, as an orderedArray

attachmentType
required
string

The image sequence input type isIMAGE_SEQUENCE

Value: "IMAGE_SEQUENCE"
extraInfoUrls
Array of strings

Additional reference information can be url link or text, needs to be aArray

object (ImageDataRecordMetadata)

Image annotation may be used metadata information

{}

Video

attachment
required
string

video data link

attachmentType
required
string

The video data type isVIDEO

Value: "VIDEO"
extraInfoUrls
Array of strings

Additional reference information can be url link or text, needs to be aArray

object (VideoDataRecordMetadata)

Video annotation may be used metadata information

{
  • "attachmentType": "VIDEO",
  • "metadata": {
    }
}

Text

Text

attachment
required
string

text data link

attachmentType
required
string

Text data type isTEXT

Value: "TEXT"
extraInfoUrls
Array of strings

Additional reference information can be url link or text, needs to be aArray

object (TextDataRecordMetadata)

Text annotation may be used metadata information

{}

Audio

Audio

attachment
required
string

audio data link

attachmentType
required
string

The audio data type isAUDIO

Value: "AUDIO"
extraInfoUrls
Array of strings

Additional reference information can be url link or text, needs to be aArray

object (AudioDataRecordMetadata)

Audio annotation may be used metadata information

{}

The data storage path for dataset creation

Text

Image

Single camera-Single frame

Single camera-Continuous frames

Multiple cameras-Single frame

Multiple cameras-Continuous frames

Point Cloud

Single radar-Single frame

Single radar-Continuous frames

Multi radar-Single frame

Multi radar-Continuous frames

(single/multi) radar (single/multi) camera-Single frame

(single/multi) radar (single/multi) camera-Continuous frames

Audio

Video

Naming requirements

  1. Point cloud mapping annotation scene requires distinguishing between radar and camera folders
  • Radar: lidar-Radar name
  • Main radarmainlidar-Radar name
  • Camera: camera-camera name

Annotation example

Single Rectangle Task

required
string or ImageSourceEntity (object) or Array of ImageSourceEntity (objects) (ImageSource)

Single frame picture input data link

attachmentType
required
string

Single frame picture input type isIMAGE

Value: "IMAGE"
extraInfoUrls
Array of strings

Additional reference information can be url link or text, needs to be aArray

object (ImageDataRecordMetadata)

Image annotation may be used metadata information

{
  • "attachmentType": "IMAGE",
  • "metadata": {