top-left position and size to rectangle of the bbox. Args: bbox: bounding box with [x, y, w, h], x,y are top-left position of the bo

Setup bundle adjustment Parameters ---------- P : ndarray of shape (K, 3, 4) Set of reconstructed camera poses X : ndarr

Update the poses and 3D points Parameters ---------- z : ndarray of shape (7K+3J,) Optimization variable n_cameras : int

args: cfg: config_file

ratio: what percent of the dataset to load, used mostly for testing purpose pre_download: pre-download frames from manifold, this is

Jitter the input box args: box - input bounding box mode - string 'template' or 'search' indicating template or search

Get all predicitons of the given label. Return empty DataFrame if there is no predcitions with the given label.

Reads ground truth file, checks if it is well formatted, and returns the ground truth instances and the activity classes. Paramete

Reads ground truth file, checks if it is well formatted, and returns the ground truth instances and the activity classes. Paramete

Converts input ground-truth to appropriate format.

Converts input ground-truth to appropriate format.

Converts input ground-truth to appropriate format.

Converts input ground-truth to appropriate format.

Reads prediction file, checks if it is well formatted, and returns the prediction instances. Parameters ----------

Converts input predictions to appropriate format.

Converts input predictions to appropriate format.

Converts input predictions to appropriate format.

Converts input predictions to appropriate format.

Performs hard-negative mining before applying the standard BCE loss. Args: same as :meth: `losses()` Returns: Dic

Performs hard-negative mining before applying the standard BCE loss. Args: same as :meth: `losses()` Returns: Dic

Helper function to register visual query datasets.

Enumerate a set of anchors for each scale wrt an anchor.