hub-visualization

name: hub-visualization description: "Use for visualization nodes in dora. Triggers on: dora-rerun, opencv-plot, rerun, visualization, plot, display, image display, bounding box visualization, 3D visualization, depth visualization, 可视化, 显示, 绘图" globs: ["/dataflow.yml", "/dataflow.yaml"] source: "https://github.com/dora-rs/dora-hub"

Visualization Nodes

Rerun and OpenCV visualization for images, 3D data, and telemetry

Available Visualization Nodes

dora-rerun

Powerful visualization using Rerun with 12 supported primitives.

Installation

pip install dora-rerun

YAML Configuration

- id: rerun
  build: pip install dora-rerun
  path: dora-rerun
  inputs:
    camera_feed: camera/image
    detections: yolo/bbox
    depth_sensor: realsense/depth
  env:
    IMAGE_WIDTH: 640
    IMAGE_HEIGHT: 480
    RERUN_MEMORY_LIMIT: 25%
    README: |
      # My Visualization
      Description shown in Rerun

Primitive-Based Visualization (v0.24.0+)

All inputs require a primitive metadata field to specify visualization type.

Method 1: Explicit Metadata in YAML

inputs:
  front_camera:
    source: camera/image
    metadata:
      primitive: "image"
  depth_sensor:
    source: realsense/depth
    metadata:
      primitive: "depth"
      focal: [600, 600]
      camera_position: [0, 0, 0]

Method 2: Sender Includes Primitive

# Sender node includes primitive in metadata
node.send_output("image", image_data, {
    "width": 640,
    "height": 480,
    "encoding": "bgr8",
    "primitive": "image"
})

Supported Primitives

Primitive Details

image

metadata = {
    "primitive": "image",
    "width": 640,
    "height": 480,
    "encoding": "bgr8"  # bgr8, rgb8, jpeg, png, avif
}

depth (with 3D reconstruction)

metadata = {
    "primitive": "depth",
    "width": 640,
    "height": 480,
    "camera_position": [0, 0, 0],        # [x, y, z]
    "camera_orientation": [0, 0, 0, 1],  # [qx, qy, qz, qw]
    "focal": [600, 600],                 # [fx, fy]
    "principal_point": [320, 240]        # [cx, cy] optional
}

boxes2d

bbox = {
    "bbox": np.array([x1,y1,x2,y2,...]).flatten(),
    "conf": np.array([0.95,...]),
    "labels": np.array(["person",...])
}
metadata = {"primitive": "boxes2d", "format": "xyxy"}

boxes3d

# center_half_size format (default)
boxes = np.array([cx,cy,cz, hx,hy,hz, ...])
metadata = {
    "primitive": "boxes3d",
    "format": "center_half_size",  # or "center_size", "min_max"
    "solid": False,  # wireframe (default) or solid
    "color": [255, 0, 0]  # RGB
}

points3d

points = np.array([x1,y1,z1, x2,y2,z2, ...], dtype=np.float32)
metadata = {
    "primitive": "points3d",
    "color": [0, 255, 0],
    "radii": [0.01, 0.01, ...]
}

lines3d

# Line segments: pairs of xyz points
lines = np.array([x1,y1,z1, x2,y2,z2, ...], dtype=np.float32)
metadata = {
    "primitive": "lines3d",
    "color": [0, 0, 255],
    "radius": 0.005
}

pose

# 7 values: position + quaternion
pose = np.array([x, y, z, qx, qy, qz, qw], dtype=np.float32)
metadata = {"primitive": "pose"}

jointstate (with URDF)

- id: rerun
  path: dora-rerun
  inputs:
    jointstate_robot: arm/joint_state
  env:
    robot_urdf: /path/to/robot.urdf
    robot_transform: "0 0.3 0"  # x y z offset

Complete Visualization Example

nodes:
  - id: camera
    build: pip install opencv-video-capture
    path: opencv-video-capture
    inputs:
      tick: dora/timer/millis/33
    outputs:
      - image

  - id: yolo
    build: pip install dora-yolo
    path: dora-yolo
    inputs:
      image: camera/image
    outputs:
      - bbox

  - id: vlm
    build: pip install dora-qwen2-5-vl
    path: dora-qwen2-5-vl
    inputs:
      image:
        source: camera/image
        queue_size: 1
    outputs:
      - text
    env:
      DEFAULT_QUESTION: "Describe the scene."

  - id: rerun
    build: pip install dora-rerun
    path: dora-rerun
    inputs:
      camera:
        source: camera/image
        metadata:
          primitive: "image"
      detections:
        source: yolo/bbox
        metadata:
          primitive: "boxes2d"
      description:
        source: vlm/text
        metadata:
          primitive: "text"
    env:
      IMAGE_WIDTH: 640
      IMAGE_HEIGHT: 480
      RERUN_MEMORY_LIMIT: 25%

opencv-plot

Simple OpenCV visualization with image, bounding boxes, and text overlay.

YAML Configuration

- id: plot
  build: pip install opencv-plot
  path: opencv-plot
  inputs:
    image: camera/image
    bbox: yolo/bbox
    text: vlm/text
  env:
    PLOT_WIDTH: 640   # Optional, defaults to image width
    PLOT_HEIGHT: 480  # Optional, defaults to image height

Input Formats

image: UInt8Array

metadata = {"width": 640, "height": 480, "encoding": "bgr8"}

bbox: StructArray

bbox = {
    "bbox": np.array([x1,y1,x2,y2,...]),
    "conf": np.array([0.95,...]),
    "labels": np.array(["person",...])
}
metadata = {"format": "xyxy"}

text: StringArray

text = pa.array(["Detection: person 95%"])

Simple Camera + Detection Visualization

nodes:
  - id: camera
    build: pip install opencv-video-capture
    path: opencv-video-capture
    inputs:
      tick: dora/timer/millis/33
    outputs:
      - image

  - id: yolo
    build: pip install dora-yolo
    path: dora-yolo
    inputs:
      image: camera/image
    outputs:
      - bbox

  - id: plot
    build: pip install opencv-plot
    path: opencv-plot
    inputs:
      image: camera/image
      bbox: yolo/bbox

Migration to dora-rerun v0.24.0

Old Way (still works with auto-detection)

inputs:
  image: camera/image  # Primitive inferred from "image" in name

New Way (recommended)

inputs:
  front_camera:
    source: camera/image
    metadata:
      primitive: "image"

Environment Variables

Related Skills

• hub-camera - Camera input nodes
• hub-detection - Detection output visualization
• hub-robot - Joint state visualization