vision_face · Face recognition

Finds faces, analyzes age, gender and emotion, and learns faces to recognize who they are.

Class: putTextPIL() Face

openpibo.vision_face.putTextPIL(img, text, points, size=30, colors=(255, 255, 255))

Writes text on an image. (Korean/English supported - uses pillow)- COPY

Parameters:

• img (numpy.ndarray) – Image object

• text (str) – Text to display

• points (tuple(int, int)) – Top-left coordinates of the text block (x, y)

• size (int) – Font size of the text

• colors (tuple(int, int, int)) – Text color BGR value (b, g, r) or hex value ‘#rrggbb’

class openpibo.vision_face.Face

Bases: object

Functions: detect() get_ageGender() get_age() get_gender() init_db() train_face() delete_face() recognize() get_db() save_db() load_db()

A class that performs various face-related functions. It can do the following.

• Detects faces.

• Learns/saves/deletes faces.

• Recognizes learned faces.

• Estimates age/gender/emotion from a face.

Face database:

Refers to the instance variable facedb, where face data is registered, loaded and deleted.

The face database format is a nested list [[], []]; the first list holds the face names and the second list holds the encoded learned face data.

It can also be saved to a file so the face information is kept after the instance is deleted.

example:

from openpibo.vision_detect import Face

face = Face()
# 아래의 모든 예제 이전에 위 코드를 먼저 사용합니다.

detect_face(img)

Detects faces.

example:

img = camera.read()
face.detect_face(img)

Parameters:

img (numpy.ndarray) – Image object

Returns:

Array of (x, y, w, h) of the detected faces. It is a list; if several faces are detected in one image, the coordinates of all detected faces are included. example:: [(10, 10, 40, 50), (120, 30, 160, 70), (130, 140, 200, 260)]

detect_face_vis(img, items)

Draws face boxes.

Parameters:

• img (numpy.ndarray) – Image object

• item (array) – List of face coordinates (x1,y1,x2,y2)

landmark_face(img, item)

Detects facial landmarks.

example:

img = camera.read()
face.landmark_face(img)

Parameters:

• img (numpy.ndarray) – Image object

• item (array) – Face coordinates (x1,y1,x2,y2)

Returns:

List of coordinates

landmark_face_vis(img, coords)

Detects facial landmarks.

example:

img = camera.read()
face.landmark_face(img)

Parameters:

• img (numpy.ndarray) – Image object

• item (array) – Face coordinates (x1,y1,x2,y2)

Returns:

List of coordinates

analyze_face(img, item)

Estimates the age, gender and emotion of a face.

example:

img = camera.read()
items = face.detect_face(img)
item = items[0] # item은 items 중 하나
face.analyze_face(img, item)

Parameters:

• img (numpy.ndarray) – Image object

• item (numpy.ndarray) – Face coordinates (x, y, w, h)

Returns:

{age: 0~100, gender: Male or Female, emotions: neutral, happy, sad, surprise, anger, box:coordinates}

analyze_face_vis(img, item)

Estimates the age, gender and emotion of a face.

Parameters:

• img (numpy.ndarray) – Image object

• item (numpy.ndarray) – Face analysis result

init_db()

Initializes the face database.

The initialized database is an empty nested list [[], []].

example:

face.init_db()

train_face(img, item, name)

Learns a face and saves it to the face database.

example:

img = camera.read()
items = face.detect_face(img)
item = items[0] # item는 items중 하나
face.train_face(img, item, 'honggildong')

Parameters:

• img (numpy.ndarray) – Image object

• item (numpy.ndarray) – Top-left and bottom-right points of the detected face rectangle (x1, y1, x2, y2)

• name (str) – Name to give the detected face

delete_face(name)

Deletes a registered face.

example:

face.delete_face('honggildong')

Parameters:

name (str) – Name of the face to delete

Returns:

True / False

recognize(img, item)

Recognizes registered faces.

example:

img = camera.read()
items = face.detect_face(img)
item = items[0] # item는 items중 하나
face.recognize(img, item)

Parameters:

• img (numpy.ndarray) – Image object

• item (numpy.ndarray) – Face coordinates (x, y, w, h)

Returns:

{"name": name, "score": error} The more similar the faces, the lower the error. Faces with an error of 0.4 or less are judged to be the same person.

get_db()

Checks the face database in use.

example:

face.get_db()

Returns:

facedb (type list(list, list)) example:: [ [‘honggildong’], [array([-0.06423206, 0.12474005, 0.0511112 , -0.05676335, -0.07211345, -0.03123881, -0.04119622, -0.12800875, 0.11717855, -0.11079554, 0.22952782, -0.02007426, -0.17457265, -0.13562854, -0.04972655, 0.15810637, -0.12785575, -0.16479518, -0.07002968, -0.00208595, 0.169218 , 0.03144928, -0.01074579, 0.04103286, -0.09245337, … -0.00706697, 0.06025593, -0.0049719 ])] ]

save_db(filename)

Saves the face database to a file.

example:

face.save_db('/home/pi/facedb')

Parameters:

filename (str) – Path of the face database file to save.

load_db(filename)

Loads a face database file.

example:

face.load_db('/home/pi/facedb')

Parameters:

filename (str) – Path of the facedb file to load.

calculate_head_orientation(keypoints)

(Internal function) Calculates face direction from facial landmark data

get_angle_between_lines(start, point1, point2)

Calculate angle between two lines defined by three points.

detect_mesh_vis(image, items)

Draw connections between landmarks based on Mediapipe’s face mesh.

detect_mesh(image)

Detect mesh and return distance, direction, and image with landmarks.