Skip to content

EdgeFirst Messages

Box

center_x class-attribute instance-attribute

center_x: float32 = 0

Normalized x-coordinate of the center

center_y class-attribute instance-attribute

center_y: float32 = 0

Normalized y-coordinate of the center

distance class-attribute instance-attribute

distance: float32 = 0

Distance of object (if known)

height class-attribute instance-attribute

height: float32 = 0

Normalized height of the box

label class-attribute instance-attribute

label: str = ''

object label

score class-attribute instance-attribute

score: float32 = 0

confidence score for detection

speed class-attribute instance-attribute

speed: float32 = 0

Speed of object (if known)

track class-attribute instance-attribute

track: Track = default_field(Track)

object tracking, each track includes ID and lifetime information

width class-attribute instance-attribute

width: float32 = 0

Normalized width of the box

Detect

boxes class-attribute instance-attribute

boxes: sequence[Box] = default_field([])

Array of detected object bounding boxes

header class-attribute instance-attribute

header: Header = default_field(Header)

Metadata including timestamp and coordinate frame

input_timestamp class-attribute instance-attribute

input_timestamp: Time = default_field(Time)

Timestamp of the input data (e.g., from camera)

model_time class-attribute instance-attribute

model_time: Time = default_field(Time)

Timestamp when the object was processed by the model

output_time class-attribute instance-attribute

output_time: Time = default_field(Time)

Timestamp when the processed output was available

DmaBuffer

fd class-attribute instance-attribute

fd: int32 = 0

The file descriptor of the DMA buffer

fourcc class-attribute instance-attribute

fourcc: uint32 = 0

The fourcc code of the image

header class-attribute instance-attribute

header: Header = default_field(Header)

Metadata including timestamp and coordinate frame

height class-attribute instance-attribute

height: uint32 = 0

The height of the image in pixels

length class-attribute instance-attribute

length: uint32 = 0

The length of the DMA buffer in bytes, used to mmap the buffer

pid class-attribute instance-attribute

pid: uint32 = 0

The process id of the service that created the DMA buffer

stride class-attribute instance-attribute

stride: uint32 = 0

The stride of the image in bytes

width class-attribute instance-attribute

width: uint32 = 0

The width of the image in pixels

Mask

encoding class-attribute instance-attribute

encoding: str = ''

The optional encoding for the mask (currently unused).

height class-attribute instance-attribute

height: uint32 = 0

The height of the mask, 0 if this dimension is unused.

length class-attribute instance-attribute

length: uint32 = 0

The length of the mask, 0 if this dimension is unused. The length would be used in 3D masks to represent the depth. It could also be used for 2D bird's eye view masks along with width instead of height (elevation).

mask class-attribute instance-attribute

mask: sequence[uint8] = default_field([])

The segmentation mask data. The array should be reshaped according to the height, width, and length dimensions. The dimension order is row-major.

width class-attribute instance-attribute

width: uint32 = 0

The width of the mask, 0 if this dimension is unused.

Model

boxes class-attribute instance-attribute

boxes: sequence[Box] = default_field([])

Array of detected object bounding boxes.

decode_time class-attribute instance-attribute

decode_time: Duration = Duration()

Duration to decode the outputs from the model, including nms and tracking.

header class-attribute instance-attribute

header: Header = Header()

Metadata including timestamp and coordinate frame

input_time class-attribute instance-attribute

input_time: Duration = Duration()

Duration to load inputs into the model

mask class-attribute instance-attribute

mask: sequence[Mask] = default_field([])

Segmentation masks from the model. Empty array if model does not generate masks. Generally models will only generate a single mask if they do.

model_time class-attribute instance-attribute

model_time: Duration = Duration()

Duration to run the model, not including input/output/decoding

output_time class-attribute instance-attribute

output_time: Duration = Duration()

Duration to read outputs from the model

RadarCube

The RadarCube interface carries various radar cube reprensentations of the Radar FFT before generally being processed by CFAR into a point cloud. The cube coud be R, RD, RAD, RA, and so on where R=Range, D=Dopper, and A=Azimuth.

Dimensional labels are used to describe the radar cube layout. Not all cubes include every label. Undefined is used for dimensions not covered by this list.

cube class-attribute instance-attribute

cube: sequence[int16] = default_field([])

The radar cube data as 16bit integers. If the is_complex is true then each element will be pairs of integers with the first being real and the second being imaginary.

header class-attribute instance-attribute

header: Header = default_field(Header)

Message header containing the timestamp and frame id.

is_complex class-attribute instance-attribute

is_complex: bool = False

True if the radar cube is complex in which case the final dimension will be doubled in size to account for the pair of int16 elements representing [real,imaginary].

layout class-attribute instance-attribute

layout: sequence[uint8] = default_field([])

Radar cube layout provides labels for each dimensions

scales class-attribute instance-attribute

scales: sequence[float32] = default_field([])

The scaling factors for the dimensions representing bins. For dimensions taken "as-is" the scale will be 1.0.

shape class-attribute instance-attribute

shape: sequence[uint16] = default_field([])

Radar cube shape provides the shape of each dimensions

timestamp class-attribute instance-attribute

timestamp: uint64 = 0

Radar frame timestamp generated on the radar module

RadarInfo

The RadarInfo interface carries the current radar configuration and status.

center_frequency class-attribute instance-attribute

center_frequency: str = ''

Radar center frequency band.

cube class-attribute instance-attribute

cube: bool = False

True if the radar is configured to output radar cubes.

detection_sensitivity class-attribute instance-attribute

detection_sensitivity: str = ''

The detection sensitivity controls the sensitivity to recognize a target.

frequency_sweep class-attribute instance-attribute

frequency_sweep: str = ''

The frequency sweep controls the detection range of the radar.

header class-attribute instance-attribute

header: Header = Header()

Message header containing the timestamp and frame id.

range_toggle class-attribute instance-attribute

range_toggle: str = ''

The range-toggle mode allows the radar to alternate between various frequency sweep configurations. Applications must handle range toggling as targets are not consistent between messages as the frequency alternates.

Track

created class-attribute instance-attribute

created: Time = default_field(Time)

Time the track was first added

id class-attribute instance-attribute

id: str = ''

Unique identifier for the object track, empty if the object is not tracked.

lifetime class-attribute instance-attribute

lifetime: int32 = 0

Number of consecutive frames the object has been tracked