-
Notifications
You must be signed in to change notification settings - Fork 0
/
distract_3dresnet.py
456 lines (383 loc) · 13.6 KB
/
distract_3dresnet.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
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
import os
import sys
import time
import warnings
import numpy as np
from PIL import Image
from PyQt5.QtWidgets import *
from PyQt5.QtGui import *
import PyQt5.QtGui as QtGui
from PyQt5.QtCore import *
import torch
import glob
import torch.nn as nn
import torch.nn.functional as F
from torchvision.transforms import transforms
from torch.utils.data import DataLoader
from torch.autograd import Variable
import torchvision
import pathlib
import cv2
import user_set
import CNN as cnn
from FacePose_pytorch.dectect import AntiSpoofPredict
from FacePose_pytorch.pfld.pfld import PFLDInference, AuxiliaryNet
from FacePose_pytorch.compute import find_pose, get_num
import PIL
from PIL import Image, ImageOps
from resnet_3d_old.opts import parse_opts
from resnet_3d_old.mean import get_mean, get_std
from resnet_3d_old.model_c import generate_model
from resnet_3d_old.spatial_transforms_winbus import (
Compose, Normalize, RandomHorizontalFlip, ToTensor, RandomVerticalFlip,
ColorAugment)
from resnet_3d_old.temporal_transforms import LoopPadding, TemporalRandomCrop, TemporalCenterCrop
# from resnet_3d_old.dataset import get_test_set
warnings.filterwarnings('ignore')
class Qt(QWidget):
def mv_Chooser(self):
opt = QFileDialog.Options()
opt |= QFileDialog.DontUseNativeDialog
fileUrl = QFileDialog.getOpenFileName(self,"Input Video", "./","Mp4 (*.mp4)", options=opt)
return fileUrl[0]
classes = cnn.read_classes('classes.txt')
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
def headpose_status(yaw, pitch, roll):
up_down = ''
left_right = ''
tilt = ''
if(yaw > user_set.H_R):
left_right = 'right'
elif(yaw < user_set.H_L):
left_right = 'left'
else:
left_right = 'normal'
if(pitch > user_set.H_D):
up_down = 'down'
elif(pitch < user_set.H_U):
up_down = 'up'
else:
up_down = 'normal'
if(roll > user_set.T_L):
tilt = 'left'
elif(roll < user_set.T_R):
tilt = 'right'
else:
tilt = 'normal'
return left_right, up_down, tilt
count = 0
yaw_sum = np.zeros(3)
yaw_count = np.zeros(3)
pitch_sum = np.zeros(3)
pitch_count = np.zeros(3)
roll_sum = np.zeros(3)
roll_count = np.zeros(3)
deg_past = np.zeros(3)
def headpose_series(yaw, pitch, roll):
global yaw_sum, yaw_count, pitch_sum, pitch_count, roll_sum, roll_count
if(abs(yaw - deg_past[0])<8):
#yaw
if(yaw>user_set.H_R):
yaw_sum[0] = yaw_sum[0] + yaw
yaw_count[0] = yaw_count[0] +1
elif(yaw<user_set.H_L):
yaw_sum[2] = yaw_sum[2] + yaw
yaw_count[2] = yaw_count[2] +1
deg_past[0] = yaw
if(abs(pitch - deg_past[1])<8):
#pitch
if(pitch>user_set.H_D):
pitch_sum[0] = pitch_sum[0] + pitch
pitch_count[0] = pitch_count[0] +1
elif(pitch<user_set.H_U):
pitch_sum[2] = pitch_sum[2] + pitch
pitch_count[2] = pitch_count[2] +1
deg_past[1] = pitch
if(abs(roll - deg_past[2])<8):
#roll
if(roll>user_set.T_L):
roll_sum[0] = roll_sum[0] + roll
roll_count[0] = roll_count[0] +1
elif(roll<user_set.T_R):
roll_sum[2] = roll_sum[2] + roll
roll_count[2] = roll_count[2] +1
deg_past[2] = roll
def headpose_output():
left_right = ''
up_down = ''
tilt = ''
if(yaw_count[0] > yaw_count[2] and yaw_sum[0]/yaw_count[0] >10):
left_right = "right"
elif(yaw_count[0] < yaw_count[2] and yaw_sum[2]/yaw_count[2] < -10):
left_right = "left"
else:
left_right = "normal"
if(pitch_count[0] > pitch_count[2] and pitch_sum[0]/pitch_count[0] >10):
up_down = "down"
elif(pitch_count[0] < pitch_count[2] and pitch_sum[2]/pitch_count[2] < -10):
up_down = "up"
else:
up_down = "normal"
if(roll_count[0] > roll_count[2] and roll_sum[0]/roll_count[0] >10):
tilt = "left"
elif(roll_count[0] < roll_count[2] and roll_sum[2]/roll_count[2] < -10):
tilt = "right"
else:
tilt = "normal"
return left_right, up_down, tilt
def dis_head(dis_status, lr, ud, ti):
score = 0
score_d = 0
score_lr = 0
score_ud = 0
score_ti = 0
if(dis_status != 'safe'):
score_d = 40
if(score_lr != 'normal' and (yaw_count[0] > 7 or yaw_count[2] > 7)):
score_lr = 40
elif(score_lr != 'normal' and (yaw_count[0] > 3 or yaw_count[2] > 3)):
score_lr = 30
elif(score_lr != 'normal' and (yaw_count[0] > 1 or yaw_count[2] > 1)):
score_lr = 10
if(score_ud != 'normal' and (pitch_count[0] > 7 or pitch_count[2] > 7)):
score_ud = 40
elif(score_ud != 'normal' and (pitch_count[0] > 3 or pitch_count[2] > 3)):
score_ud = 30
elif(score_ud != 'normal' and (pitch_count[0] > 1 or pitch_count[2] > 1)):
score_ud = 10
if(score_ti != 'normal' and (roll_count[0] > 7 or roll_count[2] > 7)):
score_ti = 30
elif(score_ti != 'normal' and (roll_count[0] > 3 or roll_count[2] > 3)):
score_ti = 20
elif(score_ti != 'normal' and (roll_count[0] > 1 or roll_count[2] > 1)):
score_ti = 10
score = score_d + np.sqrt(score_lr*score_lr + score_ud*score_ud) + score_ti
return score
def crop_range(x1, x2, y1, y2, w, h):
size = int(max([w, h]))
cx = x1 + w/2
cy = y1 + h/2
x1 = int(cx - size/2)
x2 = int(x1 + size)
y1 = int(cy - size/2)
y2 = int(y1 + size)
dx = max(0, -x1)
dy = max(0, -y1)
x1 = max(0, x1)
y1 = max(0, y1)
edx = max(0, x2 - width)
edy = max(0, y2 - height)
x2 = min(width, x2)
y2 = min(height, y2)
return x1, x2, y1, y2, dx, dy, edx, edy
def letterbox(img, resize_size, mode='square'):
shape = [img.size[1],img.size[0]] # current shape [height, width]
new_shape = resize_size
if isinstance(new_shape, int):
ratio = float(new_shape) / max(shape)
else:
ratio = max(new_shape) / max(shape) # ratio = new / old
ratiow, ratioh = ratio, ratio
new_unpad = (int(round(shape[1] * ratio)), int(round(shape[0] * ratio)))
if mode == 'auto': # minimum rectangle
dw = np.mod(new_shape - new_unpad[0], 32) / 2 # width padding
dh = np.mod(new_shape - new_unpad[1], 32) / 2 # height padding
elif mode == 'square': # square
dw = (new_shape - new_unpad[0]) / 2 # width padding
dh = (new_shape - new_unpad[1]) / 2 # height padding
elif mode == 'rect': # square
dw = (new_shape[1] - new_unpad[0]) / 2 # width padding
dh = (new_shape[0] - new_unpad[1]) / 2 # height padding
elif mode == 'scaleFill':
dw, dh = 0.0, 0.0
new_unpad = (new_shape, new_shape)
ratiow, ratioh = new_shape / shape[1], new_shape / shape[0]
top, bottom = int(round(dh - 0.1)), int(round(dh + 0.1))
left, right = int(round(dw - 0.1)), int(round(dw + 0.1))
img = img.resize(new_unpad,PIL.Image.ANTIALIAS)
img = ImageOps.expand(img, border=(left,top,right,bottom), fill=(128,128,128))
return img
class letter_img(transforms.Resize):
def __init__(self, size, interpolation=Image.BILINEAR):
assert isinstance(size, int) or (isinstance(size, Iterable) and len(size) == 2)
self.size = size
self.interpolation = interpolation
def __call__(self, img):
return letterbox(img, self.size)
def __repr__(self):
interpolate_str = _pil_interpolation_to_str[self.interpolation]
return self.__class__.__name__ + '(size={0}, interpolation={1})'.format(self.size, interpolate_str)
def get_test_data(images, spatial_transform):
images = [cv2.cvtColor(img, cv2.COLOR_BGR2RGB) for img in images]
# pre_src = cv2.cvtColor(frame,cv2.COLOR_BGR2RGB)
clip = [Image.fromarray(img) for img in images]
clip = [img.convert('RGB') for img in clip]
spatial_transform.randomize_parameters()
clip = [spatial_transform(img) for img in clip]
clip = torch.stack(clip, 0).permute(1, 0, 2, 3)
clip = torch.stack((clip,), 0)
'''
test_loader = torch.utils.data.DataLoader(
clip,
batch_size=1,
shuffle=False,
num_workers=opt.n_threads,
pin_memory=True)
'''
return clip
def predict(model, test_data):
inputs = Variable(test_data, volatile=True).cuda()
outputs = model(inputs)
outputs = F.softmax(outputs)
#print(classes[outputs.argmax()])
return classes[outputs.argmax()]
if __name__ == '__main__':
print(classes)
qt_env = QApplication(sys.argv)
process = Qt()
fileUrl = process.mv_Chooser()
print(fileUrl)
if(fileUrl == ""):
print("Without input file!!")
sys.exit(0)
left_right = ""
up_down = ""
tilt = ""
distract_output = ""
full_clip = []
distract_score = 0
output_check = np.zeros(len(classes))
font = cv2.FONT_HERSHEY_SIMPLEX
# ----------------------------------------------------
# 3D_resnet for distract detection
opt = parse_opts()
opt.mean = get_mean(opt.norm_value, dataset=opt.mean_dataset)
opt.std = get_std(opt.norm_value)
if opt.no_mean_norm and not opt.std_norm:
norm_method = Normalize([0, 0, 0], [1, 1, 1])
elif not opt.std_norm:
print('mean:', opt.mean)
norm_method = Normalize(opt.mean, [1, 1, 1])
else:
norm_method = Normalize(opt.mean, opt.std)
opt.arch = '{}-{}'.format(opt.model, opt.model_depth)
model, parameters = generate_model(opt)
checkpoint = torch.load(opt.resume_path)
opt.arch == checkpoint['arch']
model.load_state_dict(checkpoint['state_dict'])
spatial_transform = Compose([
letter_img(opt.sample_size),
#letter_img(112),
ToTensor(opt.norm_value),
norm_method
])
temporal_transform = TemporalCenterCrop(opt.sample_duration)
# -------------------------------------------------------------------------------
#model for face detect
face_model = AntiSpoofPredict(0)
#model for landmarks
headpose_model = './FacePose_pytorch/checkpoint/snapshot/checkpoint.pth.tar'
checkpoint_h = torch.load(headpose_model, map_location=device)
plfd_backbone = PFLDInference().to(device)
plfd_backbone.load_state_dict(checkpoint_h['plfd_backbone'])
plfd_backbone.eval()
plfd_backbone = plfd_backbone.to(device)
headpose_transformer = transforms.Compose([transforms.ToTensor()])
cap = cv2.VideoCapture(fileUrl)
ret, frame = cap.read()
frame = cv2.rotate(frame, cv2.cv2.ROTATE_90_CLOCKWISE)
height, width = frame.shape[:2]
fps = cap.get(cv2.CAP_PROP_FPS)
videoWriter = cv2.VideoWriter("./result.avi",cv2.VideoWriter_fourcc('X','V','I','D'),fps,(720,1280))
model.eval()
while(ret):
start=time.time()
ret, frame = cap.read()
if(not ret):
break
frame = cv2.rotate(frame, cv2.cv2.ROTATE_90_CLOCKWISE)
draw_mat = frame.copy()
# 尋找臉部範圍資訊
image_bbox = face_model.get_bbox(frame)
face_x1 = image_bbox[0]
face_y1 = image_bbox[1]
face_x2 = image_bbox[0] + image_bbox[2]
face_y2 = image_bbox[1] + image_bbox[3]
face_w = face_x2 - face_x1
face_h = face_y2 - face_y1
#尋找特徵點
crop_x1, crop_x2, crop_y1, crop_y2, dx, dy, edx, edy = crop_range(face_x1, face_x2, face_y1, face_y2, face_w, face_h)
cropped = frame[int(crop_y1):int(crop_y2), int(crop_x1):int(crop_x2)]
if (dx > 0 or dy > 0 or edx > 0 or edy > 0):
cropped = cv2.copyMakeBorder(cropped, int(dy), int(edy), int(dx), int(edx), cv2.BORDER_CONSTANT, 0)
ratio_w = face_w / 112
ratio_h = face_h / 112
cropped = cv2.resize(cropped, (112, 112))
face_input = cropped.copy()
face_input = cv2.cvtColor(face_input, cv2.COLOR_BGR2RGB)
face_input = headpose_transformer(face_input).unsqueeze(0).to(device)
_, landmarks = plfd_backbone(face_input)
pre_landmark = landmarks[0]
pre_landmark = pre_landmark.cpu().detach().numpy().reshape(-1, 2) * [112, 112]
#頭部姿態
h_start = time.time()
point_dict = {}
i = 0
for (x,y) in pre_landmark.astype(np.float32):
point_dict[f'{i}'] = [x,y]
#cv2.circle(draw_mat,(int(face_x1 + x * ratio_w),int(face_y1 + y * ratio_h)), 2, (255, 0, 0), -1)
i += 1
yaw, pitch, roll = find_pose(point_dict)
#分心狀態偵測
frame = cv2.resize(frame, (224,224))
#frame = cv2.cvtColor(frame,cv2.COLOR_BGR2GRAY)
full_clip.append(frame)
if len(full_clip) > 9:
test_data = get_test_data(full_clip, spatial_transform)
distract_output = predict(model, test_data)
full_clip = []
#頭部姿態綜合分析
if(count < 9):
headpose_series(yaw, pitch, roll)
count = count +1
else:
distract_score = 0
left_right, up_down, tilt = headpose_output()
if(face_w < 20 or face_h < 20):
left_right, up_down, tilt = "", "", ""
count = 0
distract_score = dis_head(distract_output, left_right, up_down, tilt)
output_check = np.zeros(len(classes))
yaw_sum = np.zeros(3)
yaw_count = np.zeros(3)
pitch_sum = np.zeros(3)
pitch_count = np.zeros(3)
roll_sum = np.zeros(3)
roll_count = np.zeros(3)
print(left_right, up_down, tilt)
end=time.time()
cv2.rectangle(draw_mat, (0, 0), (width, 150), (255, 255, 255), -1, cv2.LINE_AA)
cv2.rectangle(draw_mat, (face_x1, face_y1), (face_x2, face_y2), (255, 0, 255), 2, cv2.LINE_AA)
cv2.putText(draw_mat,"Head Pose ",(200,35), font,1.1,(0,0,0),2)
cv2.putText(draw_mat,"LEFT_RIGHT",(200,70), font,0.8,(255,0,0),2)
cv2.putText(draw_mat,"UP_DOWN ",(200,100), font,0.8,(255,0,0),2)
cv2.putText(draw_mat,"TILT ",(200,130), font,0.8,(255,0,0),2)
cv2.putText(draw_mat, ": "+str(left_right), (360,70), font,0.8,(255,0,0),2)
cv2.putText(draw_mat, ": "+str(up_down), (360,100), font,0.8,(255,0,0),2)
cv2.putText(draw_mat, ": "+str(tilt), (360,130), font,0.8,(255,0,0),2)
cv2.putText(draw_mat,"Status",(15,35), font,1.1,(0,0,0),2)
cv2.putText(draw_mat,distract_output,(15,100), font, 1,(0,0,255),2)
cv2.putText(draw_mat,"FPS : "+str(int(1/(end-start))),(550,135), font, 0.8,(0,0,0),2)
if(distract_score >= 30):
cv2.rectangle(draw_mat, (500, 10), (width-5, 100), (120, 0, 255), 2, cv2.LINE_AA)
cv2.putText(draw_mat,"dangerous!",(510,60), font, 1.1,(120,0,255),2,cv2.LINE_AA)
#draw_mat = cv2.resize(draw_mat,(360,640))
#print((1/(end-start)))
cv2.imshow("frame", draw_mat)
videoWriter.write(draw_mat)
if cv2.waitKey(1) & 0xFF == ord('q'):
cap.release()
cv2.destroyAllWindows()
break
videoWriter.release()
cap.release()