Data Structures

This section present the 3 internal data structures implemented in inaFaceAnalyzer.

• inaFaceAnalyzer.rect.Rect : a rectangle implementation usefull for manipulating face bounding boxes

• inaFaceAnalyzer.face_detector.Detection : obtained from face detection systems, containing face bounding boxes together with face detection confidence

• inaFaceAnalyzer.face_tracking.TrackDetection : an extension of Detection used when combining face detection and tracking

The remaining data structures used to exchange data between modules are based on pandas.DataFrame, allowing easy exports to various table formats (see pandas’s documentation).

namedtuple inaFaceAnalyzer.rect.Rect(x1: float, y1: float, x2: float, y2: float)

Bases: NamedTuple

This class is an internal data structure allowing to manipulate rectangle shapes

Fields:

0.  x1 (float) – left

1.  y1 (float) – top

2.  x2 (float) – right

3.  y2 (float) – bottom

__contains__(point)

point contained in self

Parameters:

point (tuple) – (x,y)

Returns:

True if point is in self, else False

Return type:

bool

property area

self Surface area

property center

center (x,y) of self (x1, y1, x2, y2)

static from_dlib(x)

create Rect from dlib’s rectangle instance

Parameters:

x (dlib.rectangle or dlib.drectangle) –

Return type:

Rect

property h

self height

intersect(r)

Rectangle intersection between self and r

Parameters:

r (Rect) –

Return type:

Rect

iou(r)

Intersection Over Union between self and r

Parameters:

r (Rect) –

Return type:

float

property max_dim_len

max (width, height)

mult(x, y)

Multiply self coordinates by horizontal and vertical scaling factors Usefull for converting [0…1] coordinates to image frame dimensions in pixels

Parameters:

• (float (x) – horizontal scaling factor.

• y (float) – vertical scaling factor.

Return type:

Rect

scale(scale_prct)

scale self according to a given scale percentage

Parameters:

scale_prct (float) –

Return type:

Rect

property square

returns the smallest square containing the rectangle

to_dlibFloat()

Convert self to dlib.drectangle (float)

Return type:

dlib.drectangle

to_dlibInt()

Convert self to dlib.rectangle (int)

Return type:

dlib.rectangle

to_int()

Convert self coordinates (float) to the nearest int values :rtype: Rect

transpose(xoffset, yoffset)

Translation

Parameters:

• xoffset (float) – horizontal offset.

• yoffset (float) – vertical offset.

Return type:

Rect

property w

self width

namedtuple inaFaceAnalyzer.face_detector.Detection(bbox: Rect, detect_conf: float)

Bases: NamedTuple

Atomic element returned by face detection classes

Fields:

0.  bbox (Rect) – position of the detected face in the image in pixels

1.  detect_conf (float) – face detection confidence (0 = lowest confidence, 1 = highest confidence)

namedtuple inaFaceAnalyzer.face_tracking.TrackDetection(bbox: Rect, face_id: int, detect_conf: float, track_conf: float)

Bases: NamedTuple

Atomic element returned by face tracking and detection classes

Fields:

0.  bbox (Rect) – bounding box

1.  face_id (int) – detected face numerical identifier

2.  detect_conf (float) – face detection confidence

3.  track_conf (float) – tracking confidence