Get a parameter value with the given name. If it does not exists, it return the default value given as a second argument or returns an error i

shape - (width, height) tuple

Samples num_ids frames between min_id and max_id for which target is visible args: visible - 1d Tensor indicating whether target

Update the state of the particle filter given an observation. Parameters: ---------- observed: array The observe

The input type is NestedTensor, which consists of: - tensor: batched images, of shape [batch_size x 3 x H x W] - mask: a binary mask o

Convert a quaternion to rotation matrix Parameters ---------- q : ndarray of shape (4,) Unit quaternion (w, x, y, z)

Register a new instance and add it to the InstanceRegistration object. If the instance ID is not in the register, add a new ID and O

Obtain the nearest frame for a given time, video fps, and feature window.

Loads a network checkpoint file. Can be called in three different ways: load_checkpoint(): Loads the latest epoch

Freezes backbone, proposal_generator. Sets roi_heads to train.

Create configs and perform basic setups.

Convert a rotation matrix to quaternion Parameters ---------- R : ndarray of shape (3, 3) Rotation matrix Returns

Extracts region from a bounding box in the image with some context padding. Arguments: image - (1, c, h, w) Tensor bbox - bo

Given dataset paths and other configuration arguments, it registers a visual query dataset with the specified arguments.

Extracts a square crop centered at target_bb box, of area search_area_factor^2 times target_bb area args: im - cv image target_bb

get soft-argmax coordinate for a given heatmap

get soft-argmax coordinate for a given heatmap

Evaluate the reprojection error Parameters ---------- z : ndarray of shape (7K+3J,) Optimization variable b : ndarray of

Return an activation function given a string

Same as `torch.nn.functional.binary_cross_entropy_with_logits`, but returns 0 (instead of nan) for empty inputs. `flip_class` inverts the tar

Concatenates two dicts without copying internal data.

:param corner: Corner or np.array 4*N :return: Center or 4 np.array N

Given a set of clips from a video, estimate how much of the video is effectively covered by the clips. Args: video_duration - Le

NOTICE: search region related features must be in the last place

Old style model is stored with all names of parameters share common prefix 'module.

To reset the tracker after done running on one video. Clear any intermediate variables to get a clean setup for next video.

Compute the temporal intersection over union between a target segment and all the test segments. Parameters ---------- target_segment

Computes spatio-temporal IoU between a target response track (prediction) and multiple candidate response tracks (ground-truth).

Do training for the given number of epochs. args: max_epochs - Max number of training epochs, load_latest - Bool indic

Top-left x, y to center x, y. Args: bbox: bounding box with [x, y, w, h], x,y are top-left position of the box. context_rati