-
Notifications
You must be signed in to change notification settings - Fork 0
/
generator_utils.py
84 lines (73 loc) · 2.82 KB
/
generator_utils.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
import os
import shutil
from tqdm import tqdm
import cv2
import numpy as np
import pickle
face_detector = cv2.CascadeClassifier('haarcascade_frontalface_default.xml')
IMAGE_SIZE = 96
def find_face(img):
gray = img.copy()
# gray=cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
faces = face_detector.detectMultiScale(gray, 1.3, 5)
for (x, y, w, h) in faces:
face = gray[y:y + h, x:x + w]
face = cv2.resize(face, (IMAGE_SIZE, IMAGE_SIZE))
return True, face
return False, gray
def cleanImages(findFace=True):
for folder in os.listdir('Dataset'):
folder = os.path.join('Dataset', folder)
if len(os.listdir(folder)) < 3:
shutil.rmtree(folder)
continue
if findFace:
for img_path in os.listdir(folder):
img_path = os.path.join(folder, img_path)
img = cv2.imread(img_path)
flag, face = find_face(img)
os.remove(img_path)
if flag:
cv2.imwrite(img_path, face)
#cleanImages(findFace=True)
cleanImages(findFace=False)
input_shape = (3, IMAGE_SIZE, IMAGE_SIZE)
faces = []
images = {}
folders = os.listdir('Dataset')
length = len(folders)
length = 10
for folder_indx in tqdm(range(length)):
folder = os.path.join('Dataset', folders[folder_indx])
li = []
key = ''
for img in os.listdir(folder):
key = img
img1 = cv2.imread(os.path.join(folder, img))
#img2 = img1[..., ::-1]
cv2.imshow('img1',img1)
cv2.imshow('img',np.around(img1 / 255.0, decimals=12))
cv2.waitKey(0)
li.append(np.around(np.transpose(img2, (2, 0, 1)) / 255.0, decimals=12))
images[key] = np.array(li)
faces.append(key)
def batch_generator(batch_size=16):
y_val = np.zeros((batch_size, 2, 1))
anchors = np.zeros((batch_size, input_shape[0], input_shape[1], input_shape[2]))
positives = np.zeros((batch_size, input_shape[0], input_shape[1], input_shape[2]))
negatives = np.zeros((batch_size, input_shape[0], input_shape[1], input_shape[2]))
TOTAL_SIZE = 0
while True:
for i in range(batch_size):
positiveFace = faces[np.random.randint(len(faces))]
negativeFace = faces[np.random.randint(len(faces))]
while positiveFace == negativeFace:
negativeFace = faces[np.random.randint(len(faces))]
positives[i] = images[positiveFace][np.random.randint(len(images[positiveFace]))]
anchors[i] = images[positiveFace][np.random.randint(len(images[positiveFace]))]
negatives[i] = images[negativeFace][np.random.randint(len(images[negativeFace]))]
x_data = {'anchor': anchors,
'anchorPositive': positives,
'anchorNegative': negatives
}
yield (x_data, [y_val, y_val, y_val])