keypoints_processing

Description

Util functions to process, transform, and evaluate MS COCO keypoint information.

The main functionality is:

• annotation transformation: pad_and_scale_annotation(), annotations_to_mask()

• estimation of the size (in pixels) of a person in an image from an annotation: annotation_to_sizes(), annotation_to_tot_height()

• some functions that may serve as filtering conditions: person_has_rel_size(), all_keypts_visible(), any_keypts_visible()

Attributes

COCO_STD_KEYPOINT_NAMES

The standard list of keypoint names for MS COCO.

Functions

all_keypts_visible(ann, keypoint_idxs)	Whether all given keypoints are marked as visible in the given annotation.
annotation_to_sizes(annotation[, ...])	Estimate the body size of a person in an image in pixels from skeletal keypoints and linear formulas given by factors.
annotation_to_tot_height(annotation[, ...])	Wrapper around annotation_to_sizes() that returns the cleaned maximum total height.
annotations_to_heatmap(annotation_wh, ...[, ...])	Create a heatmap marking the centroids of the linked keypoints from the annotations list.
annotations_to_mask(*, annotation_wh, ...[, ...])	Create a mask of the linked keypoints from the annotations list.
any_keypts_visible(ann[, keypoint_idxs])	Whether any of the given keypoints are marked as visible in the given annotation.
pad_and_scale_annotation(ann, from_size[, ...])	Given a MS COCO style annotation, apply first padding and/or then scaling to its coordinates.
person_has_rel_size(img_meta, ann[, ...])	Check whether the relative size of the person described by annotation ann wrt image size is in specified range.
to_keypoint_idxs(keypoint_names[, coco, ...])	Return the start indices for the given keypoints within an annotations 'keypoint' field.