Camera Component

A camera component is a source of 2D and/or 3D images. You can use the component to configure a webcam, lidar, time-of-flight sensor, or another type of camera.

The API for camera components allows you to:

  • Request single images or a stream in 2D color, or display z-depth.

  • Request a point cloud. Each 3D point cloud image consists of a set of coordinates (x,y,z) representing depth in mm.

The configuration of your camera component depends on your camera model. You can use different models to:

  • Configure physical cameras that generate images or point clouds.
  • Combine streams from multiple cameras into one.
  • Transform and process images.

Supported models

To use your camera with Viam, check whether one of the following built-in models or modular resources supports your camera.

Built-in models

For configuration information, click on the model name:

ModelDescription
ffmpegUses a camera, a video file, or a stream as a camera.
image_fileGets color and depth images frames from a file path.
velodyneUses velodyne lidar.
webcamA standard camera that streams camera data.
fakeA camera model for testing.
join_color_depthJoins the outputs of a color and depth camera already registered in your config to create a third “camera” that outputs the combined and aligned image.
join_pointcloudsCombines the point clouds from multiple camera sources and projects them to be from the point of view of target_frame.
transformA pipeline for applying transformations to an input image source.
ultrasonicThe HC-S204 ultrasonic distance sensor.

Modular resources

Search for additional camera models that you can add from the Viam Registry:

For configuration information, click on the model name:

Model
Description

Control your camera with Viam’s client SDK libraries

To get started using Viam’s SDKs to connect to and control your machine, go to your machine’s page on the Viam app, navigate to the CONNECT tab’s Code sample page, select your preferred programming language, and copy the sample code generated.

When executed, this sample code will create a connection to your machine as a client. Then control your machine programmatically by adding API method calls as shown in the following examples.

These examples assume you have a camera called "my_camera" configured as a component of your machine. If your camera has a different name, change the name in the code.

Be sure to import the camera package for the SDK you are using:

from viam.components.camera import Camera
import (
  "go.viam.com/rdk/components/camera"
)

API

The camera component supports the following methods:

Method NameDescription
GetImageReturn an image from the camera.
GetImagesReturn multiple simultaneous images from different imagers, along with associated metadata.
GetPointCloudReturn a point cloud from the camera.
GetPropertiesReturn the camera intrinsic and camera distortion parameters, as well as whether the camera supports returning point clouds.
GetGeometriesGet all the geometries associated with the camera in its current configuration, in the frame of the camera.
DoCommandSend or receive model-specific commands.
CloseSafely shut down the resource and prevent further use.

GetImage

Returns an image from the camera. You can request a specific MIME type but the returned MIME type is not guaranteed. If the server does not know how to return the specified MIME type, the server returns the image in another format instead.

Parameters:

  • mime_type (str): The requested MIME type of the image. The returned MIME type is not guaranteed to match the image output type: If the requested MIME type has +lazy appended to it or the requested MIME type is not supported for decoding/encoding, GetImage returns a RawImage instead.

Returns:

my_camera = Camera.from_robot(robot=robot, name="my_camera")

frame = await my_camera.get_image(mime_type="image/jpeg")

If the mime_type of your image is image/vnd.viam.dep, pass the returned image data to the Viam Python SDK’s RawImage.bytes_to_depth_array() method to decode the raw image data to a standard 2D image representation.

For example:

# Assume "frame" has a mime_type of "image/vnd.viam.dep"
frame = await my_camera.get_image()

# Convert "frame" to a standard 2D image representation.
# Remove the 1st 3x8 bytes and reshape the raw bytes to List[List[Int]].
standard_frame = frame.bytes_to_depth_array()

For more information, see the Python SDK Docs.

Parameters:

  • ctx (Context): A Context carries a deadline, a cancellation signal, and other values across API boundaries.
  • errHandlers (ErrorHandler): A handler for errors allowing for logic based on consecutively retrieved errors).

Returns:

myCamera, err := camera.FromRobot(robot, "my_camera")

// gets the stream from a camera
stream, err := myCamera.Stream(context.Background())

// gets an image from the camera stream
img, release, err := stream.Next(context.Background())
defer release()

For more information, see the Go SDK Docs.

GetImages

Get simultaneous images from different imagers, along with associated metadata. The multiple images returned from GetImages() do not represent a time series of images.

Parameters:

  • None.

Returns:

my_camera = Camera.from_robot(robot=robot, name="my_camera")

images, metadata = await my_camera.get_images()
img0 = images[0].image
timestamp = metadata.captured_at

For more information, see the Python SDK Docs.

Parameters:

  • ctx (Context): A Context carries a deadline, a cancellation signal, and other values across API boundaries.

Returns:

myCamera, err := camera.FromRobot(robot, "my_camera")

images, metadata, err := myCamera.Images(context.Background())

For more information, see the Go SDK Docs.

GetPointCloud

Get a point cloud from the camera as bytes with a MIME type describing the structure of the data. The consumer of this call should decode the bytes into the format suggested by the MIME type.

Parameters:

  • None.

Returns:

  • (Tuple[bytes,str]): The pointcloud data as bytes paired with a string representing the MIME type of the pointcloud (for example, PCD).

To deserialize the returned information into a numpy array, use the Open3D library:

import numpy as np
import open3d as o3d

my_camera = Camera.from_robot(robot=robot, name="my_camera")

data, _ = await my_camera.get_point_cloud()

# write the point cloud into a temporary file
with open("/tmp/pointcloud_data.pcd", "wb") as f:
    f.write(data)
pcd = o3d.io.read_point_cloud("/tmp/pointcloud_data.pcd")
points = np.asarray(pcd.points)

For more information, see the Python SDK Docs.

Parameters:

  • ctx (Context): A Context carries a deadline, a cancellation signal, and other values across API boundaries.

Returns:

  • (pointcloud.PointCloud): A general purpose container of points. It does not dictate whether or not the cloud is sparse or dense.
  • (error): An error, if one occurred.
myCamera, err := camera.FromRobot(robot, "my_camera")

pointCloud, err := myCamera.NextPointCloud(context.Background())

For more information, see the Go SDK Docs.

GetProperties

Get the camera intrinsic parameters and camera distortion, as well as whether the camera supports returning point clouds.

Parameters:

  • None.

Returns:

my_camera = Camera.from_robot(robot=robot, name="my_camera")

properties = await my_camera.get_properties()

For more information, see the Python SDK Docs.

Parameters:

  • ctx (Context): A Context carries a deadline, a cancellation signal, and other values across API boundaries.

Returns:

  • (Properties): Properties of the particular implementation of a camera.
  • (error): An error, if one occurred.
myCamera, err := camera.FromRobot(robot, "my_camera")

// gets the properties from a camera
properties, err := myCamera.Properties(context.Background())

For more information, see the Go SDK Docs.

GetGeometries

Get all the geometries associated with the camera in its current configuration, in the frame of the camera. The motion and navigation services use the relative position of inherent geometries to configured geometries representing obstacles for collision detection and obstacle avoidance while motion planning.

Parameters:

  • extra (Optional[Dict[str, Any]]): Extra options to pass to the underlying RPC call.
  • timeout (Optional[float]): An option to set how long to wait (in seconds) before calling a time-out and closing the underlying RPC call.

Returns:

For more information, see the Python SDK Docs.

my_camera = Camera.from_robot(robot=robot, name="my_camera")

geometries = await my_camera.get_geometries()

if geometries:
    # Get the center of the first geometry
    print(f"Pose of the first geometry's centerpoint: {geometries[0].center}")

DoCommand

Execute model-specific commands that are not otherwise defined by the component API. For native models, model-specific commands are covered with each model’s documentation. If you are implementing your own camera and adding features that have no native API method, you can access them with DoCommand.

Parameters:

Returns:

my_camera = Camera.from_robot(robot, "my_camera")

command = {"cmd": "test", "data1": 500}
result = my_camera.do(command)

For more information, see the Python SDK Docs.

Parameters:

Returns:

myCamera, err := camera.FromRobot(robot, "my_camera")

command := map[string]interface{}{"cmd": "test", "data1": 500}
result, err := myCamera.DoCommand(context.Background(), command)

For more information, see the Go SDK Code.

Close

Safely shut down the resource and prevent further use.

Parameters:

  • None

Returns:

  • None
my_camera = Camera.from_robot(robot, "my_camera")

await my_camera.close()

For more information, see the Python SDK Docs.

Parameters:

  • ctx (Context): A Context carries a deadline, a cancellation signal, and other values across API boundaries.

Returns:

  • (error) : An error, if one occurred.
myCamera, err := camera.FromRobot(robot, "my_camera")

err := myCamera.Close(ctx)

For more information, see the Go SDK Docs.

Next steps


You can also ask questions in the Community Discord and we will be happy to help.