Args: param1: Integer parameter. param2: Float parameter. param3: String

Register subclass in the tracker registry if it defines tracker_id. Extracts parameter metadata from __init__ at class definition time.

Call BoTSORTTracker._get_associated_indices with default similarity threshold.

Call ByteTrackTracker._get_associated_indices with the given matrix.

Call OCSORTTracker._get_associated_indices with zeros for direction matrix.

Call SORTTracker._get_associated_indices with n_tracks dummy tracks.

Whether the source has a known total frame count.

Sample hyperparameters, run tracker over all sequences, return metric. Args: trial: Optuna trial used to sample parameter values.

Independent BIoU reference: buffer boxes, then apply vanilla IoU.

Reference CIoU from torchvision.

Reference DIoU from torchvision.

Reference GIoU from torchvision.

Transform points from absolute (world) to relative (frame) coordinates. Args: points: Array of shape `(N, 2)` containing `(x, y)`

Transform from absolute (world) to relative (frame) coordinates.

Apply CMC to all active tracks. .. deprecated:: 2.5 Use CMC.apply_batch(H, self.tracks) directly. Args: H: 2

Deprecated alias for :meth:`CMC.warp_xyxy_corners`. .. deprecated:: 2.5 Renamed to :meth:`CMC.warp_xyxy_corners`. This wrapper fo

A 40x60 bounding box in xyxy format.

Clear the trajectory for a specific tracker ID. Args: tracker_id: The tracker ID to clear.

Create a state estimator for the given state representation. Args: state_repr: The desired representation. Ex: StateRepresentation.XCYCSR

Fixture providing DanceTrack flat format test data.

Fixture providing DanceTrack MOT17 format test data.

Directory with valid images followed by one corrupted image file.

Directory containing only non-image files.

Empty directory with no files.

Create `CLEARMetrics` from a dictionary. Args: data: Dictionary with metric values. Returns: `CLEARMetrics`

Create `IdentityMetrics` from a dictionary. Args: data: Dictionary with metric values. Returns: `IdentityMet

Create `SequenceResult` from a dictionary. Args: data: Dictionary with sequence name and metrics. Returns: `

Create `BenchmarkResult` from a dictionary. Args: data: Dictionary with sequences and aggregate results. Returns:

Snapshot the filter state for later restoration (e.g. ORU freeze). Returns: Opaque state dictionary.

Get current bounding box estimate from the filter/state.

Parsed args with only required flags.

Normalize similarity values for score fusion in BoT-SORT association. By default returns the matrix unchanged. Signed variants (GIoU, DIoU,

Process page HTML content to remove doctest directives. This hook runs after markdown is converted to HTML, so we need to handle HTML-en

Build TechArticle + speakable JSON-LD for the page and store in page.meta.

Run the Kalman filter prediction step.

Predict next bounding box position and advance missed-frame state. Propagates the Kalman filter and increments `time_since_update` (and

Predict next state (prior) using state transition model. Computes: x = F @ x P = F @ P @ F.T + Q

Return True if user pressed quit key.

Transform points from relative (frame) to absolute (world) coordinates. Args: points: Array of shape `(N, 2)` containing `(x, y)`

Transform from relative (frame) to absolute (world) coordinates.

Clear all stored trajectories. Call this when switching videos or when you want to reset trajectory history.

Clear all internal tracking state. Call between videos or when tracking should restart from scratch.

Reset tracker state by clearing all tracks and resetting ID counter. Call this method when switching to a new video or scene.

Reset the estimator state. Call this when starting a new video or when you want to reset the world coordinate system to the current f