6. Use Camera for Perception

Our Scenes come with two predefined cameras:

• rgbd - an RGB-D camera on the robot's end-effector
• rgbd_cage - an RGB-D camera in an over-the-shoulder position

Cameras are RGB-D Cameras, meaning you can also request the depth data for each image. As an alternative to RGB color, you can also use the get_segmentation() function, which will return image data as if a pixel-perfect object segmentation has been performed.

To add a new Camera to your Scene, load it as if it were any other object:

new_cam = Camera(...)
scene.add_object(new_cam)
scene.start()


new_cam.get_image(...)

Adjust the camera parameters

To adjust camera parameters like image-width and -height you can use the set_cam_params(...) method. This also allows to configure deeper camera parameters.

In some functions like calc_point_cloud we use the camera parameters to calculate three dimensional positions. It is therefore not recommended to use custom widths and sizes with these functions. You should rather set these params before with set_caram_params.

new_cam = Camera(...)
scene.add_object(new_cam)

new_cam.set_cam_params(
    width,  # the width of the camera images
    height, # the height of the camera images
    near,   # the near plane for the depth image, in the raw depth image the depth values will use `near` as 0
    far,    # the far plane for the depth image, in the raw depth image the depth values will use `near` as 1
    fovy,   # The field of view of the camera in y-direction. fovx will then be calculated based on fovy and image dimensions
)

scene.start()


new_cam.get_image(...)

6.1 ROS Camera control

TBA ... Back to Overview