Initialize layers of the transformer head.

Initialize layers of the transformer head.

Initialize layers of the transformer head.

Add extra files that are required to support MIM into the package. These files will be added by creating a symlink to the originals if the pac

forest: a set of arbitrary Tree output: a set of corresponding binary Tree

Run the detector in test mode, given gt_bboxes, return the labels, dists Return: det_labels: 1 based.

Run the detector in test mode, given gt_bboxes, return the labels, dists Return: det_labels: 1 based.

Run the detector in test mode, return the detected boxes, labels, dists, and masks.

Draws keypoints of an instance and follows the rules for keypoint connections to draw lines between appropriate keypoints. This follo

Args: box_coord (tuple): a tuple containing x0, y0, x1, y1 coordinates, where x0 and y0 are the coordinates of th

Args: circle_coord (list(int) or tuple(int)): contains the x and y coordinates of the center of the circle.

Draw annotations/segmentaions in Detectron2 Dataset format. Args: dic (dict): annotation/segmentation data of one image, in Detec

Draw panoptic prediction annotations or results. Args: panoptic_seg (Tensor): of shape (height, width) where the values are ids f

Draw a rotated box with label on its top-left corner. Args: rotated_box (tuple): a tuple containing (cnt_x, cnt_y, w, h, angle),

Draw semantic segmentation predictions/labels. Args: sem_seg (Tensor or ndarray): the segmentation of shape (H, W).

Object context and object classification. :param obj_feats: [num_obj, img_dim + object embedding0 dim] :return: edge_ctx: [num_obj, #

Object context and object classification. :param obj_feats: [num_obj, img_dim + object embedding0 dim] :return: edge_ctx: [num_obj, #

Returns: pred_to_gt: Matching from predicate to GT pred_5ples: the predicted (id0, id1, cls0, cls1, rel) pred_triplet_sco

Function: This is used for getting the results in both float data form and text form so that they can be logged into tensorboard

Forward function. Args: feats (tuple[Tensor]): Features from the upstream network, each is a 4D-tensor.

Step 1: Divide all nodes into left child container and right child container Step 2: From left child container and right child container, sele

generate a list of trees that covers all the objects in a batch det_result: pair_scores: [obj_num, obj_num] output: list of trees, ea

NOTE:Transform network outputs for a batch into psg predictions, but still use the name of get_bboxes for now. Args: all_

Equivalent to nn.functional.interpolate, but with support for empty batch sizes. This will eventually be supported natively by PyTorch, and t

Equivalent to nn.functional.interpolate, but with support for empty batch sizes. This will eventually be supported natively by PyTorch, and t

Load word vectors from a path, trying .pt, .txt, and .zip extensions.

Loss function. Only outputs from the last feature level are used for computing losses by default. Args: all_cls_

Does the hairy LSTM math. :param timestep_input: :param previous_state: :param previous_memory: :param dropout_mask:

Map rois to corresponding feature levels by scales. - scale < finest_scale * 2: level 0 - finest_scale * 2 <= scale < finest_scale *

NMS for multi-class bboxes. Args: multi_bboxes (Tensor): shape (n, #class*4) or (n, 4) multi_scores (Tensor): shape (n, #class),

NMS for multi-class bboxes. Args: multi_bboxes (np.array): shape (n, #class*4) or (n, 4) multi_scores (np.array): shape (n, ),

Object context and object classification. :param num_objs: :param obj_feats: [num_obj, img_dim + object embedding0 dim] :para

Object context and object classification. :param obj_feats: [num_obj, img_dim + object embedding0 dim] :param obj_labels: [num_obj] t

Args: boxes (ndarray): an Nx5 numpy array of (x_center, y_center, width, height, angle_degrees) format