video2.py
4.72 KB
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
import dlib
import cv2
import numpy as np
import matplotlib.pyplot as plt
import tensorflow as tf
import math
import os
import pathlib
import time
import pandas as pd
import tensorflow as tf
from tensorflow.keras.preprocessing.image import ImageDataGenerator,load_img
from tensorflow.keras.models import load_model
from tensorflow.keras import regularizers
from tensorflow import keras
from imutils import face_utils
import time
start = time.time()
detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor("./models/shape_predictor_68_face_landmarks.dat")
facerec = dlib.face_recognition_model_v1('models/dlib_face_recognition_resnet_model_v1.dat')
model = load_model('../checkpoint/er-best-mobilenet1-bt32-model-classweight-adam.h5')
# model = load_model('../checkpoint/er-best-mobilenet2-bt32-model-adam.h5')
# model = load_model('../checkpoint/er-best-efficientNet1-bt32-model-SGD.h5')
descs = np.load('img/descs2.npy', allow_pickle=True)[()]
video_path = './data/zoom_1.mp4'
cap=cv2.VideoCapture(video_path)
labels_dict_ = {0 : 'angry', 1 : 'fear' , 2: 'happy', 3: 'neutral', 4: 'sad', 5: 'surprise'}
# labels_dict_ = {'angry' : 0,'fear' : 1 ,'happy' : 2, 'neutral' : 3, 'sad' : 4, 'surprise' : 5}
time_dict = {'angry': [], 'fear': [], 'happy': [], 'neutral': [], 'sad': [], 'surprise': []}
def get_key(val):
for key, value in labels_dict_.items():
if(value == val):
return key
def convertMillis(millis):
seconds=(millis/1000)%60
minutes=(millis/(1000*60))%60
hours=(millis/(1000*60*60))%24
return seconds, int(minutes), int(hours)
#cap = cv2.VideoCapture(0) # 0번 카메라
# 동영상 크기(frame정보)를 읽어옴
frameWidth = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
frameHeight = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
frame_size = (frameWidth, frameHeight)
fps = cap.get((cv2.CAP_PROP_FPS))
_, img_bgr = cap.read() # (800, 1920, 3)
padding_size = 0
resized_width = 1920
video_size = (resized_width, int(img_bgr.shape[0] * resized_width // img_bgr.shape[1]))
timestamps = [cap.get(cv2.CAP_PROP_POS_MSEC)]
prev_time = 0
fourcc = cv2.VideoWriter_fourcc('m', 'p', '4', 'v')
# out1 = cv2.VideoWriter('./data/record0.mp4',fourcc, fps, frame_size)
while True:
retval, frameBGR = cap.read() # 영상을 한 frame씩 읽어오기
current_time = time.time() - prev_time
if(type(frameBGR) == type(None)):
pass
else:
frameBGR = cv2.resize(frameBGR, video_size)
frame = cv2.cvtColor(frameBGR, cv2.COLOR_BGR2RGB)
if (retval is True) and (current_time > 1.5) :
prev_time = time.time()
faces = detector(frame, 1)
for (i, face) in enumerate(faces):
shape = predictor(frame, face)
face_descriptor = facerec.compute_face_descriptor(frame, shape)
img = cv2.resize(frame[face.top():face.bottom(), face.left():face.right()], dsize=(224, 224), interpolation = cv2.INTER_CUBIC)
imgarr = np.array(img).reshape(1, 224, 224, 3) /255
emotion = labels_dict_[model.predict(imgarr).argmax(axis=-1)[0]]
# emotion = get_key(model.predict_classes(imgarr))
last_found = {'name': 'unknown', 'dist': 0.6, 'color': (0,0,255)}
for name, saved_desc in descs.items():
dist = np.linalg.norm([face_descriptor] - saved_desc, axis=1)
if dist < last_found['dist']:
last_found = {'name': name, 'dist': dist, 'color': (255,255,255)}
cv2.rectangle(frameBGR, pt1=(face.left(), face.top()), pt2=(face.right(), face.bottom()), color=last_found['color'], thickness=2)
cv2.putText(frameBGR, last_found['name'] + ',' + emotion , org=(face.left(), face.top()), fontFace=cv2.FONT_HERSHEY_SIMPLEX, fontScale=1, color=last_found['color'], thickness=2)
# cv2.putText(frameBGR, last_found['name'] + ',' , org=(face.left(), face.top()), fontFace=cv2.FONT_HERSHEY_SIMPLEX, fontScale=1, color=last_found['color'], thickness=2)
con_sec, con_min, con_hour = convertMillis(cap.get(cv2.CAP_PROP_POS_MSEC))
time_dict[emotion].append("{0}:{1}:{2}".format(con_hour, con_min, round(con_sec, 3)))
print("{0}:{1}:{2} {3}".format(con_hour, con_min, round(con_sec, 3), emotion))
# print("{0}:{1}:{2} {3}".format(con_hour, con_min, con_sec))
cv2.imshow('frame', frameBGR)
key = cv2.waitKey(25)
if key == 27 :
break
print(time_dict)
print("총 시간 : ", time.time() - start)
if cap.isOpened():
cap.release()
for i in range(1,5):
cv2.destroyAllWindows()
cv2.waitKey(1)