Kinect CHOP

From TouchDesigner 099 Wiki


The Kinect CHOP reads positional and skeletal tracking data from the Kinect and Linect2 sensors. Up to 6 people's full skeletons can be tracked (2 on the original Kinect), and the center position of an addition 4 people in the camera view is tracked as well. Multiple Kinect devices can be used by using multiple Kinect CHOPs and changing the Camera parameter.

It also supports face tracking.

NOTE: This CHOP works with the Kinect for Windows hardware and the Kinect 2, not Kinect for Xbox hardware.

To use a Kinect 2 device you need to install the Kinect 2 SDK or runtime from here.

To use the original Kinect 1 you will need to install the Kinect Runtime 1.8 for builds greater than 12000, or Kinect Runtime 1.7 for builds below 12000.

See also Kinect TOP, Kinect and Kinect1.

PythonIcon.png kinectCHOP_Class

Parameters - Kinect Page

Hardware Version hwversion - Choose between Kinect v1 or Kinect v2 sensors.

Sensor sensor - Selects which Kinect sensor to use. Only available when using Kinect v1

Skeleton skeleton - Selects options for skeletal tracking.

  • Full - Track full skeleton.
  • Seated - Track seated skeleton.
  • Off - Do not track skeleton.

Max Players maxplayers - Limits how many players to track.

Interactions interactions - Enables interactions, which returns interaction data such as "grip" and "press". The additional channels output are prefixed with p[1-2]/hand_l_* and p[1-2]/hand_r_*.

  • For the x and y values of the interaction hand, 0.0 corresponds to left edge of interaction region and 1.0 corresponds to right edge of interaction region, but values could be outside of this range.
  • For the z value of the interaction hand, 0.0 means that hand is close to shoulder and 1.0 represents a fully extended arm, but values could be outside of this range.

NOTE: Interaction mode requires a depth resolution of 640x480. As such, a Kinect TOP displaying a depth image for the same sensor will output an image with a resolution of 640x480 and the Near Depth Mode setting will be set in the Kinect CHOP.

For guidelines in creating applications with Kinect Interaction, download the Kinect for Windows Human Interface Guidelines.

Relative Bone Orientations relboneorientations - Returns rx, ry, and rz relative rotation channels for each bone. The rotation is relative to the previous joint.

Absolute Bone Orientations absboneorientations - Returns rx, ry, and rz absolute rotation channels for each bone.

Bone Lengths bonelengths - Returns the length for each bone.

Unroll Bone Values unrollbones - Return the rx, ry, and rz values in a form that includes no discontinuities (or wrap arounds), so that they can be safely blended, filtered, lagged etc. For example, a rotation value running from 0 to 360 continuously, unrolled would start at 0 and increase in value indefinitely.

World Space Positions worldspace - Returns the tracked positions in world space coordinates. For each tracked point, a tx, ty, tz triplet of channels will be output.

Image Space Positions imagespace - Returns the tracked positions in image space coordinates. On the Kinect 2 this is relative to the color image, on the Kinect 1 it is relative to the depth image. For each tracked point, a u, v, tz triplet of channels will be output.

Face Tracking facetracking - Returns face tracking channels for detected faces.

Status Channels statuschans - A number of additional status channels are reported when this is turned on. A group of channels will report if a joint is currently being tracked, and another group of channels reports if part of the player is clipped outside the bounds of the Kinect sensor. The additional channels output are suffixed with tracked or clipped.

Smoothing smoothing - Increasing the smoothing parameter value leads to more highly-smoothed skeleton position values being returned. It is the nature of smoothing that, as the smoothing value is increased, responsiveness to the raw data decreases. Thus, increased smoothing leads to increased latency in the returned skeleton values. Values must be in the range 0 through 1.0. Passing 0 causes the raw data to be returned.

Correction correction - Lower values are slower to correct towards the raw data and appear smoother, while higher values will correct toward the raw data more quickly. Values must be in the range 0 through 1.0.

Prediction prediction - The number of frames to predict into the future. Values must be greater than or equal to zero. Values greater than 0.5 will likely lead to overshooting when moving quickly. This effect can be damped by using small values of Max Deviation Radius.

Jitter Radius jitterrad - The radius in meters for jitter reduction. Any jitter beyond this radius is clamped to the radius.

Max Deviation Radius maxdevrad - The maximum radius in meters that filtered positions are allowed to deviate from raw data. Filtered values that would be more than this radius from the raw data are clamped at this distance, in the direction of the filtered value.

For more information about smoothing, see the Joint Smoothing White Paper.

Near Depth Mode neardepthmode - Enables near mode for skeleton tracking, which allows the depth camera to see objects as close as 40cm to the camera (instead of the default 80cm).

Standard Options and Local Variables


Device Support

This CHOP only works with the Kinect for Windows hardware device. Kinect for Xbox360 devices will not work.

See Also