-
Notifications
You must be signed in to change notification settings - Fork 56
Design
The most basic set of tasks to fully produce the expected output. In this case, reading images from one or more cameras, to a set of tracker poses sent to a driver.
- Blocking
Blocks until an image is ready, synced to FPS of camera.
- Depends: Capture Image
Convert to single channel.
- Blocking
Get a tracker pose from the BridgeDriver for the previous frame. Once smoothing and multi-cam are moved to ATT, this won't be needed.
- Depends: Post-Process Image, Get Tracker Pose
Using last frame's detections, black out areas of the image that a tracker is not expected to be found.
- Depends: Mask Image Search Area
- Slow
Calls apriltag_detector_detect on the image and returns a list of detections.
- Depends: Capture Image, Calculate Tracker Pose
- Depends: Detect Tags In Image,
- Depends: Calculate Tracker Pose
- Blocking
Position and rotation of an object.
Unique pattern for computer vision recognition, with enough information/calibration, pose can be calculated. AKA fiducial tag.
Fiducial tag specification for a set of patterns, and implementation for detecting tags in an image. One of the most accurate tags for realtime applications. github.
Provides pose to a VR application, could be vive tracker, or in our case, one or more tags.
One or more tags found in an image. Made of the 2d pixel coordinates of the corners and center, and the unique Id.