Figment

Detect Pose

Detect one or more human poses in the input image and draw them as skeletons on a solid background. The node runs the MediaPipe pose detector and pose landmark models on the GPU through ONNX Runtime; the MediaPipe library itself is not used. Only the landmark coordinates come back, 33 per person.

Connect the landmarks output to a Send OSC node to send the landmarks to another application. To cut people out of the image instead of drawing them, use Segment Pose.

Parameters

  • Background The color that fills the output image behind the drawing.
  • Coloring solid draws every point and line in the colors below. per limb gives each landmark and each limb a fixed color, so the left and right sides and every limb are distinguishable by hue. Use per limb when the drawing is the input of an image-to-image model such as pix2pix: the model can then tell which limb is which. The point and line colors are ignored in this mode. The palette is borrowed from OpenPose, so the look is familiar; the drawing is not a drop-in replacement for an OpenPose render.
  • Draw Points Draw each landmark as a dot.
  • Color (points) The color of the dots.
  • Radius The size of the dots.
  • Draw Lines Connect the landmarks of each pose with lines.
  • Color (lines) The color of the lines.
  • Line Width The thickness of the lines.
  • Number of Poses The maximum number of people to detect, from 1 to 4. Each extra person costs one more landmark pass.
  • Model The size of the landmark model: lite, full, or heavy. Larger models are more accurate and slower. Switching the model reloads it; the node keeps running the current model until the new one is ready.
  • Mode video follows each person from frame to frame and only runs the detector again when a person is lost. This is faster and steadier for a webcam or movie. still runs the detector on every frame. Use it for unrelated images, such as a Load Image Folder.
  • Smoothing Filters the landmarks over time in video mode so they stop jittering. 0 is off. Higher values are steadier but add a little lag on fast movement; the filter follows fast motion more closely than slow motion, so around 0.65 (the strength MediaPipe uses) is a good starting point. The drawn skeleton and the Landmarks output are both smoothed. Has no effect in still mode.

Outputs

  • Out The drawing on the background color, at the size of the input image.
  • Detected true while at least one person is in the frame.
  • Landmarks An object with type: 'pose' and a landmarks array with one entry per person. Each entry is a list of points with x, y, and z normalized to the image (0 to 1) and a visibility score. The output is null when nobody is detected.