Accepts a list of spans and the the corresponding caption. Returns a cleaned list of spans where: - Overlapping spans are merged -

Computes the area of a set of bounding boxes, which are specified by its (x1, y1, x2, y2) coordinates. Args: boxes (Tensor[N, 4]

Generalized IoU from https://giou.stanford.edu/ The boxes should be in [x0, y0, x1, y1] format Returns a [N, M] pairwise matrix, where

Compute and display summary metrics for evaluation results.

The function names in programs and templates don't always match 1:1, so we redefine equality

The function names in programs and templates don't always match 1:1, so we redefine equality

Return intersection-over-union (Jaccard index) of boxes. Both sets of boxes are expected to be in ``(x1, y1, x2, y2)`` format with ``0 <

Given a node idx in the given program, try to find the corresponding node in the template

Get ann ids that satisfy given filter conditions. Args: img_ids (int array): get anns for given imgs cat_ids (int arra

Given a sentence, returns a tuple with the following items: -- root text:str : the text associated with the root of the sentence -- negative_

Load anns with the specified ids. If ids=None load all anns. Args: ids (int array): integer array of annotation ids Return

Convert [x y w h] box format to [x1 y1 x2 y2] format.

Accumulate per image evaluation results and store the result in self.eval.

:return: [x_min, y_min, x_max, y_max] of all boxes

Given a set of boxes, returns the average of all clusters of boxes that are highly overlapping. A high overlap is characterized by the iou_thresho

Do the heavy lifting on the given subset (eg 'train')

construct a map such that positive_map[i,j] = True iff box i is associated to token j

Given a node idx in the given program, try to find the corresponding node in the template

Returns: True if distributed training is enabled

Load categories with the specified ids. If ids=None load all images. Args: ids (int array): integer array of image ids Ret

Returns True if the given spans intersect

Moves the target dicts to the given device.

Class for reading and visualizing annotations. Args: annotation_path (str): location of annotation file

Return an activation function given a string

Return an activation function given a string

Create gt, dt which are list of anns/dets. If use_cats is true only anns/dets corresponding to tuple (img_id, cat_id) will be used. El

Prepare self._gts and self._dts for evaluation based on params.

Log a negative hit @k for given category

Log a positive hit @k for given category

Run all_gather on arbitrary picklable data (not necessarily tensors) Args: data: any picklable object Returns: list[data]

Convert annotation which can be polygons, uncompressed RLE to RLE. Args: ann (dict) : annotation object Returns:

Find previous filtering nodes that might have text associated. The full filter may contain UNION nodes, which means that there might be severa