-
Notifications
You must be signed in to change notification settings - Fork 154
/
TYY_create_db_biwi.py
205 lines (166 loc) · 5.7 KB
/
TYY_create_db_biwi.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
import scipy.io as sio
import pandas as pd
from os import listdir
from os.path import isfile, join
from tqdm import tqdm
import sys
import cv2
from moviepy.editor import *
import numpy as np
import argparse
from mtcnn.mtcnn import MTCNN
def get_args():
parser = argparse.ArgumentParser(description="This script cleans-up noisy labels "
"and creates database for training.",
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument("--db", type=str, default='./BIWI',
help="path to database")
parser.add_argument("--output", type=str, default='./BIWI_noTrack.npz',
help="path to output database mat file")
parser.add_argument("--img_size", type=int, default=64,
help="output image size")
parser.add_argument("--ad", type=float, default=0.4,
help="enlarge margin")
args = parser.parse_args()
return args
def main():
args = get_args()
mypath = args.db
output_path = args.output
img_size = args.img_size
ad = args.ad
isPlot = True
detector = MTCNN()
onlyfiles_png = []
onlyfiles_txt = []
for num in range(0,24):
if num<9:
mypath_obj = mypath+'/0'+str(num+1)
else:
mypath_obj = mypath+'/'+str(num+1)
print(mypath_obj)
onlyfiles_txt_temp = [f for f in listdir(mypath_obj) if isfile(join(mypath_obj, f)) and join(mypath_obj, f).endswith('.txt')]
onlyfiles_png_temp = [f for f in listdir(mypath_obj) if isfile(join(mypath_obj, f)) and join(mypath_obj, f).endswith('.png')]
onlyfiles_txt_temp.sort()
onlyfiles_png_temp.sort()
onlyfiles_txt.append(onlyfiles_txt_temp)
onlyfiles_png.append(onlyfiles_png_temp)
print(len(onlyfiles_txt))
print(len(onlyfiles_png))
out_imgs = []
out_poses = []
for i in range(len(onlyfiles_png)):
print('object %d' %i)
mypath_obj = ''
if i<9:
mypath_obj = mypath+'/0'+str(i+1)
else:
mypath_obj = mypath+'/'+str(i+1)
for j in tqdm(range(len(onlyfiles_png[i]))):
img_name = onlyfiles_png[i][j]
txt_name = onlyfiles_txt[i][j]
img_name_split = img_name.split('_')
txt_name_split = txt_name.split('_')
if img_name_split[1] != txt_name_split[1]:
print('Mismatched!')
sys.exit()
pose_path = mypath_obj+'/'+txt_name
# Load pose in degrees
pose_annot = open(pose_path, 'r')
R = []
for line in pose_annot:
line = line.strip('\n').split(' ')
L = []
if line[0] != '':
for nb in line:
if nb == '':
continue
L.append(float(nb))
R.append(L)
R = np.array(R)
T = R[3,:]
R = R[:3,:]
pose_annot.close()
R = np.transpose(R)
roll = -np.arctan2(R[1][0], R[0][0]) * 180 / np.pi
yaw = -np.arctan2(-R[2][0], np.sqrt(R[2][1] ** 2 + R[2][2] ** 2)) * 180 / np.pi
pitch = np.arctan2(R[2][1], R[2][2]) * 180 / np.pi
img = cv2.imread(mypath_obj+'/'+img_name)
img_h = img.shape[0]
img_w = img.shape[1]
if j==0:
[xw1_pre,xw2_pre,yw1_pre,yw2_pre] = [0,0,0,0]
detected = detector.detect_faces(img)
if len(detected) > 0:
dis_list = []
XY = []
for i_d, d in enumerate(detected):
xv = []
yv = []
for key, value in d['keypoints'].items():
xv.append(value[0])
yv.append(value[1])
if d['confidence'] > 0.90:
x1,y1,w,h = d['box']
x2 = x1 + w
y2 = y1 + h
xw1 = max(int(x1 - ad * w), 0)
yw1 = max(int(y1 - ad * h), 0)
xw2 = min(int(x2 + ad * w), img_w - 1)
yw2 = min(int(y2 + ad * h), img_h - 1)
# Crop the face loosely
# x_min = int(min(xv))
# x_max = int(max(xv))
# y_min = int(min(yv))
# y_max = int(max(yv))
# h = y_max-y_min
# w = x_max-x_min
# xw1 = max(int(x_min - ad * w), 0)
# xw2 = min(int(x_max + ad * w), img_w - 1)
# yw1 = max(int(y_min - ad * h), 0)
# yw2 = min(int(y_max + ad * h), img_h - 1)
XY.append([xw1,xw2,yw1,yw2])
dis_betw_cen = np.abs(xw1-img_w*2/3)+np.abs(yw1-img_h*2/3)
dis_list.append(dis_betw_cen)
if len(dis_list)>0:
min_id = np.argmin(dis_list)
[xw1,xw2,yw1,yw2] = XY[min_id]
dis_betw_frames = np.abs(xw1-xw1_pre)
if dis_betw_frames < 80 or j==0:
img = cv2.resize(img[yw1:yw2 + 1, xw1:xw2 + 1, :], (img_size, img_size))
[xw1_pre,xw2_pre,yw1_pre,yw2_pre] = [xw1,xw2,yw1,yw2]
if isPlot:
print([xw1_pre,xw2_pre,yw1_pre,yw2_pre])
cv2.imshow('check',img)
k=cv2.waitKey(10)
img = cv2.resize(img, (img_size, img_size))
cont_labels = np.array([yaw, pitch, roll])
out_imgs.append(img)
out_poses.append(cont_labels)
# else:
# img = cv2.resize(img[yw1_pre:yw2_pre + 1, xw1_pre:xw2_pre + 1, :], (img_size, img_size))
# # Checking the cropped image
# if isPlot:
# print([xw1_pre,xw2_pre,yw1_pre,yw2_pre])
# print('Distance between two frames too large! Use previous frame detected location.')
# cv2.imshow('check',img)
# k=cv2.waitKey(10)
# img = cv2.resize(img, (img_size, img_size))
# cont_labels = np.array([yaw, pitch, roll])
# out_imgs.append(img)
# out_poses.append(cont_labels)
# else:
# img = cv2.resize(img[yw1_pre:yw2_pre + 1, xw1_pre:xw2_pre + 1, :], (img_size, img_size))
# if isPlot:
# print('No face detected! Use previous frame detected location.')
# # Checking the cropped image
# if isPlot:
# cv2.imshow('check',img)
# k=cv2.waitKey(10)
# img = cv2.resize(img, (img_size, img_size))
# cont_labels = np.array([yaw, pitch, roll])
# out_imgs.append(img)
# out_poses.append(cont_labels)
np.savez(output_path,image=np.array(out_imgs), pose=np.array(out_poses), img_size=img_size)
if __name__ == '__main__':
main()