-
Notifications
You must be signed in to change notification settings - Fork 0
/
testing_imagen_face.py
279 lines (206 loc) · 8.13 KB
/
testing_imagen_face.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
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
import pickle
import os
import configparser
import torch
import numpy as np
from transformers import AutoProcessor, ASTModel
from scipy.io import wavfile
import torch
import pickle
from pyAudioAnalysis import audioBasicIO
from pyAudioAnalysis import ShortTermFeatures
import librosa
import subprocess
import os
import pathlib
import configparser
import cv2
import random
from deepface import DeepFace
import matplotlib.pyplot as plt
import sqlite3 as sl
from PIL import Image
import numpy as np
import math
import threading
import time
from multiprocessing import Process
from threading import Thread
import pandas as pd
from math import nan
def resizeImage(im, size):
im = cv2.cvtColor(im, cv2.COLOR_BGR2RGB)
image = Image.fromarray(im.astype('uint8'), 'RGB')
x, y = image.size
if(x >= y):
x_n = size
y_n = int(math.floor((x_n/x) * y))
else:
y_n = size
x_n = int(math.floor((y_n/y) * x))
image_n = image.resize((x_n,y_n), Image.ANTIALIAS)
return image_n
def make_square(im, min_size, fill_color=(0, 0, 0, 0)):
x, y = im.size
size = max(min_size, x, y)
new_im = Image.new('RGBA', (size, size), fill_color)
new_im.paste(im, (int((size - x) / 2), int((size - y) / 2)))
return new_im
def extract_face(df_data,resizeImageTo,fddfb,output_folder,
efvr,efhr, auto_zoom = True,start_at_frame = -1):
# Loading configurations
configParser = configparser.RawConfigParser()
configFilePath = r'configuration.txt'
configParser.read(configFilePath)
insert_amd_env_vars = int(configParser.get('COMMON', 'insert_amd_env_vars'))
HSA_OVERRIDE_GFX_VERSION = configParser.get('COMMON', 'HSA_OVERRIDE_GFX_VERSION')
ROCM_PATH = configParser.get('COMMON', 'ROCM_PATH')
if(insert_amd_env_vars != 0):
os.environ["HSA_OVERRIDE_GFX_VERSION"] = HSA_OVERRIDE_GFX_VERSION
os.environ["ROCM_PATH"] = ROCM_PATH
df_data_sub= df_data.apply(extract_face_fn,args=(
start_at_frame,output_folder,resizeImageTo,fddfb,efvr,efhr,auto_zoom
),axis=1)
df_data['gender'] = df_data_sub[0]
df_data['ethnicity'] = df_data_sub[1]
df_data['age'] = df_data_sub[2]
df_data['VGG'] = df_data_sub[3]
#print(df_data)
df_data = df_data[df_data['gender'].notna()]
df_data = df_data[df_data['ethnicity'].notna()]
df_data = df_data[df_data['age'].notna()]
df_data = df_data[df_data['VGG'].notna()]
with open(output_folder + '/' + 'df_data1.pickle', 'wb') as handle:
pickle.dump(df_data, handle)
def extract_face_fn(row,start_at_frame,output_folder,resizeImageTo,fddfb,efvr,efhr,auto_zoom):
vidcap = cv2.VideoCapture(row['absPathVideo'])
framesNo = int(vidcap.get(cv2.CAP_PROP_FRAME_COUNT))
if(start_at_frame < 0):
frame_number = random.randint(1, framesNo - 1)
else:
frame_number = start_at_frame
frameNo=1
absPathFrame = output_folder + "/" + str(row['index']) + "frame.png"
vidcap.set(cv2.CAP_PROP_POS_FRAMES, frame_number-1)
success,image = vidcap.read()
count = 1
cv2.imwrite(absPathFrame, image) # save frame as PNG file
vidcap.release()
backends = [
'opencv',
'ssd',
'dlib',
'mtcnn',
'retinaface',
'mediapipe'
]
try:
face_objs = DeepFace.extract_faces(img_path = absPathFrame,
target_size = (resizeImageTo, resizeImageTo),
detector_backend = backends[fddfb]
)
except:
os.remove(absPathFrame)
s = pd.Series([nan,nan,nan,nan])
return s
if(len(face_objs) == 0):
os.remove(absPathFrame)
s = pd.Series([nan,nan,nan,nan])
return s
# Gets biggest face in image (because images can have multiple faces)
facial_area = None
size_one_facial_area_prev = 0
for face_obj in face_objs:
one_facial_area = face_obj['facial_area']
one_facial_area_h = one_facial_area['h']
one_facial_area_w = one_facial_area['w']
size_one_facial_area = one_facial_area_h * one_facial_area_w
if(size_one_facial_area > size_one_facial_area_prev):
size_one_facial_area_prev = size_one_facial_area
facial_area = one_facial_area
crop_img_start_row = facial_area['y'] - int(efvr * facial_area['h'])
if(crop_img_start_row < 0):
crop_img_start_row = 0
crop_img_start_col = facial_area['x'] - int(efhr * facial_area['w'])
if(crop_img_start_col < 0):
crop_img_start_col = 0
crop_img_end_row = facial_area['y'] + int(efvr * facial_area['h'])+facial_area['h']
if(crop_img_end_row >= len(image)):
crop_img_end_row = len(image) - 1
crop_img_end_col = facial_area['x'] + int(efhr * facial_area['w'])+facial_area['w']
if(crop_img_end_col >= len(image[0])):
crop_img_end_col = len(image[0]) -1
crop_img = image[crop_img_start_row : crop_img_end_row,
crop_img_start_col :crop_img_end_col]
crop_img = resizeImage(crop_img,resizeImageTo)
crop_img = make_square(crop_img,resizeImageTo)
if(auto_zoom):
w_s = resizeImageTo / (1+2 * 0.4)
h_s = resizeImageTo / (1+2 * 0.4)
#print(image.size)
crop_img = crop_img.crop((0.2*w_s,0.0*h_s,1.6*w_s,1.4*h_s))
crop_img = crop_img.resize((resizeImageTo,resizeImageTo))
#absPathFace = output_folder + "/" +str(index)+ "face.png"
crop_img.save(row['absPathFace'])
try:
face_analysis_objs = DeepFace.analyze(img_path = row['absPathFace'],
actions = ['age', 'gender', 'race'],enforce_detection=False)
embedding_objs = DeepFace.represent(row['absPathFace'],enforce_detection=False)
embedding = embedding_objs[0]["embedding"]
except:
os.remove(absPathFrame)
s = pd.Series([nan,nan,nan,nan])
return s
if(len(face_analysis_objs) == 1):
gender = face_analysis_objs[0]['dominant_gender']
ethnicity = face_analysis_objs[0]['dominant_race']
age = face_analysis_objs[0]['age']
else:
os.remove(absPathFrame)
s = pd.Series([nan,nan,nan,nan])
return s
os.remove(absPathFrame)
s = pd.Series([gender,ethnicity,age,embedding])
return s
def extract_face_attr_and_rep_fn(row):
#print(row['face_path'])
try:
face_analysis_objs = DeepFace.analyze(img_path = row['face_path'],
actions = ['age', 'gender', 'race'],enforce_detection=False)
embedding_objs = DeepFace.represent(row['face_path'],enforce_detection=False)
embedding = embedding_objs[0]["embedding"]
except:
s = pd.Series([nan,nan,nan,nan])
return s
if(len(face_analysis_objs) == 1):
gender = face_analysis_objs[0]['dominant_gender']
ethnicity = face_analysis_objs[0]['dominant_race']
age = face_analysis_objs[0]['age']
else:
s = pd.Series([nan,nan,nan,nan])
return s
s = pd.Series([gender,ethnicity,age,embedding])
return s
def extract_face_attr_and_rep(df_data,output_folder):
# Loading configurations
configParser = configparser.RawConfigParser()
configFilePath = r'configuration.txt'
configParser.read(configFilePath)
insert_amd_env_vars = int(configParser.get('COMMON', 'insert_amd_env_vars'))
HSA_OVERRIDE_GFX_VERSION = configParser.get('COMMON', 'HSA_OVERRIDE_GFX_VERSION')
ROCM_PATH = configParser.get('COMMON', 'ROCM_PATH')
if(insert_amd_env_vars != 0):
os.environ["HSA_OVERRIDE_GFX_VERSION"] = HSA_OVERRIDE_GFX_VERSION
os.environ["ROCM_PATH"] = ROCM_PATH
df_data_sub = df_data.apply(extract_face_attr_and_rep_fn,args=(),axis=1)
df_data['gender'] = df_data_sub[0]
df_data['ethnicity'] = df_data_sub[1]
df_data['age'] = df_data_sub[2]
df_data['face_rep'] = df_data_sub[3]
#print(df_data)
df_data = df_data[df_data['gender'].notna()]
df_data = df_data[df_data['ethnicity'].notna()]
df_data = df_data[df_data['age'].notna()]
df_data = df_data[df_data['face_rep'].notna()]
with open(output_folder + '/' + 'df_data_100.pickle', 'wb') as handle:
pickle.dump(df_data, handle)