Body Class

class plotbee.body.Body(parts, center, angle_conn, connections, frame, body_id=- 1, suppressed=False, pollen_score=0.0, tag=None, features=None, virtual=False, annotations={}, metadata={})

Bases: object

__dict__ = mappingproxy({'__module__': 'plotbee.body', 'y_offset': 0, 'width': 200, 'height': 400, 'scale': 1.0, 'out_width': None, 'out_height': None, 'cX': None, 'cY': None, 'ignore_angle': False, 'pollen_threshold': 0.5, 'valid_function': <function valid_fn>, 'load_body': <classmethod object>, '__init__': <function Body.__init__>, 'annotate': <function Body.annotate>, 'set_metadata': <function Body.set_metadata>, 'annotations': <property object>, 'metadata': <property object>, 'valid': <property object>, 'frameid': <property object>, 'tag_id': <property object>, 'video_name': <property object>, 'connections': <property object>, 'id': <property object>, 'parts': <property object>, 'center': <property object>, 'angle': <property object>, 'pollen': <property object>, '_image': <function Body._image>, 'image': <property object>, 'skeleton_image': <property object>, 'prev': <property object>, 'next': <property object>, 'set_id': <function Body.set_id>, 'cbox': <function Body.cbox>, 'boundingBox': <function Body.boundingBox>, 'skeleton': <property object>, '__len__': <function Body.__len__>, '__repr__': <function Body.__repr__>, 'save': <function Body.save>, 'info': <function Body.info>, 'tag_erased_image': <function Body.tag_erased_image>, 'tag_image': <function Body.tag_image>, 'params': <function Body.params>, '__dict__': <attribute '__dict__' of 'Body' objects>, '__weakref__': <attribute '__weakref__' of 'Body' objects>, '__doc__': None, '__annotations__': {}})
__init__(parts, center, angle_conn, connections, frame, body_id=- 1, suppressed=False, pollen_score=0.0, tag=None, features=None, virtual=False, annotations={}, metadata={})

Default Constructor

Parameters

  • parts (dict[list[tuple(x:int, y:int)]]) – detected parts

  • center (int) – part used for keypoint tracking

  • angle_conn (tuple(int, int)) – connection used to compute angle

  • connections (list[list[x:int, y:int]]) – predicted connection configuration, note this can be different for each body

  • frame (Frame) – Frame object which this body belongs

  • body_id (int, optional) – track_if if -1 the body do not have an id assigned, defaults to -1

  • suppressed (bool, optional) – True if this body is suppressed, defaults to False

  • pollen_score (float, optional) – pollen score predicted from a pollen detection model, defaults to 0.0

  • tag (dict, optional) – tag dictionary computed by apriltag, defaults to None

  • features (np.ndarray, optional) – identity features computed by a identity model, defaults to None

  • virtual (bool, optional) – True, if this body wasn not detected by beepose but detected by tracking, defaults to False

  • annotations (dict, optional) – annotations (key-value), defaults to dict()

  • metadata (dict, optional) – metadata (key-value), defaults to dict()

Note

The parts is a list of list of keypoints, if there is a case that a body has two detected keypoint of on part because a duplicate detection due to the skeleton association. Ideally, this list should contain the keypoints with more confidence first. I believe I never found a case for this, but the part attribute was left as this just in case.

__len__()

Return the number of detected parts(not duplicates)

Returns

number of detected parts

Return type

int

__module__ = 'plotbee.body'
__repr__()

Print body info

Returns

Show body id, parts points and if virtual

Return type

str

__weakref__

list of weak references to the object (if defined)

_image(width=None, height=None, cX=None, cY=None, scale=None, ignore_angle=None, erase_tag=False)

Get a body image

Parameters

  • width (int, optional) – image width, defaults to None

  • height (int, optional) – image height, defaults to None

  • cX (int, optional) – x of the center point in the output rectangle of the body image, defaults to None

  • cY (int, optional) – y of the center point in the output rectangle of the body image, defaults to None

  • scale (float, optional) – scale of the body image, defaults to None

  • ignore_angle (bool, optional) – ignore angle normalization, defaults to None

  • erase_tag (bool, optional) – if body is a tagged bee then the tag will be erase from the image, defaults to False

Returns

body image

Return type

numpy.ndarray

Note

If the following params width, height, cX, cY, scale and ignore_angle are set to None, default values will be used. These default values are the class attributes Body.`param` (e.g. Body.width).

Tip

For a batch image extraction or plotting use class attributes to change image extarction at global (Body) level.

property angle

Return angle

Returns

Return angle in degrees from the angle connection

Return type

float

annotate(key, value)

Make a body annotation

Parameters

  • key (Any hashable (ideal str)) – name of the annotation field.

  • value (Any) – Value of the annotation.

Note

Note that annotation can overwrite previous annotations.

property annotations

get annotation dictionary

Returns

annotations

Return type

dict

boundingBox()

Get keypoint of a rotated bounding box of the body

Returns

4 x-y points of the rotated bounding box.

Return type

tuple(tuple(x, y))

Note

this bounding box uses the following as parameters self.center, self.angle, Body.width and Body.height.

cX = None

Class Attribute for cX for plotting

cY = None

Class Attribute for cX for plotting

cbox(w_size=100, h_size=200)

Get keypoint of a rectangle in the center of the body

Parameters

  • w_size (int, optional) – rectangle width, defaults to 100

  • h_size (int, optional) – rectangle height, defaults to 200

Returns

rectangle points (top-left x, top-left y, bottom-right x, bottom-right y)

Return type

tuple(x1, y1, x2, y2)

property center

get body center

Returns

get center point

Return type

tuple(x, y)

Note

y = Body.center.y - Body.y_offset

property connections

Get parts connections (skeleton)

Returns

list of parts connections

Return type

list[list[x:int, y:int]]

property frameid

Get frame id

Returns

frame id

Return type

int

height = 400

Class Attribute for height for plotting

property id

Get body id

Returns

get track id

Return type

int

ignore_angle = False

Class Attribute for ignore_angle for plotting (this is used to extract images without angle normalization)

property image

Get Body image

Returns

body image

Return type

numpy.ndarray

info()

Get body info as dictionary

Returns

body info

Return type

dict

Warning

This is old. Better use self.params.

classmethod load_body(body_dict, frameobj)

Body Constructor from dictionary and Frame object

Parameters

  • body_dict (dict) – recieves a dictionary with the following:

  • frameobj (Frame) – parent frame (or future parent frame)

Returns

Body object

Return type

Body

Note

body_dict have the following keys:

  • parts

  • center_part

  • angel_conn

  • connections

  • id

  • suppressed

  • pollen_score

  • tag (optional)

  • features (optional)

  • virtual (optional)

  • annotations (optional)

  • metadata (optional)

see Body constructor for more information about these keys.

Note

Frame object is used to have a two-way link between Frame and Body. Generally, frameobj is a new and empty frame that will be populated with the data from body_dict.

property metadata

get metadata dictionary

Returns

metdatda

Return type

dict

property next

Get the next body in the track. This is only work with tracks.

Warning

Not sure id new tracking support this linked-list track.

out_height = None

Class Attribute for out_height for plotting

out_width = None

Class Attribute for out_width for plotting

params()

Get body dictionary

Returns

return a dictionary of the body.

Return type

dict

Warning

For developers. If body has new attributes, these attributes should be manually code into this function to be able save and load bodies.

property parts

Get detected parts

Returns

detected parts

Return type

dict[list[tuple(x:int, y:int)]]

property pollen

Return if the body has pollen

Returns

return if is body with pollen

Return type

bool

Note

Pollen decision is based on self.pollen_score > Body.pollen_threshold

pollen_threshold = 0.5

Class Attribute for pollen_threshold this is used as self.pollen_score > Body.pollen_threshold

property prev

Get the previous body in the track. This is only work with tracks.

Warning

Not sure id new tracking support this linked-list track.

save(path, width=None, height=None, cX=None, cY=None, erase_tag=False)

Save body image

Parameters

  • path (str) – path to save the image

  • width (int, optional) – image width, defaults to None

  • height (int, optional) – image height, defaults to None

  • cX (int, optional) – x of the center point in the output rectangle of the body image, defaults to None

  • cY (int, optional) – y of the center point in the output rectangle of the body image, defaults to None

  • scale (float, optional) – scale of the body image, defaults to None

  • ignore_angle (bool, optional) – ignore angle normalization, defaults to None

  • erase_tag (bool, optional) – if body is a tagged bee then the tag will be erase from the image, defaults to False

scale = 1.0

Class Attribute for scale for plotting

set_id(i)

Set track id (body id)

Parameters

i (int) – body id

set_metadata(key, value)

Set metadata

Parameters

  • key (Any hashable (ideal str)) – name of the metadata field.

  • value (Any) – Value of metadata.

Note

Note that set_metadata can overwrite previous metadata.

property skeleton

Get skeleton

Returns

Skeleton connections (list of lines)

Return type

list[tuple(tuple(x:int, y:int))]

property skeleton_image

Get body image with skeleton overlay

Returns

body image with skeleton on top

Return type

numpy.ndarray

suppressed

Instance attribute to suppress body. Generally, use during tracking to remove double detections.

tag_erased_image()

Get tag erased image

Returns

return body image. If body is tagged then erase the tag.

Return type

numpy.ndarray

property tag_id

Tag id

Returns

None if is the body is not tagged

Return type

float

tag_image()

Return image

Returns

If body is tagged then return image of the Apriltag else return np.array([])

Return type

np.ndarray

property valid

Return if body is a valid

Returns

check if body is a valid body with the function defined on Body.valid_function().

Return type

bool

Tip

This is customizable, and be useful to filter bodies in multiple scenarios such as plotting or Frame.bodies

valid_function()

Class Attribute (function) for to decided what is a valid body

property video_name

Get video name

Returns

video name

Return type

str

width = 200

Class Attribute for width for plotting

y_offset = 0

Class Attribute for y_offset for Body.center()

Note

This was used to centralize the body center when there is not a center keypoint available. For example, in my case I used to move the center to the thorax using the neck keypoint as Body.center.

plotbee.body.parse_parts(parts)
plotbee.body.valid_fn(body)

Default Valid function Filters suppressed bodies and bodies with id = -1.

Parameters

body (Body) – body check

Returns

True if body is valid else False

Return type

bool