/
recognition_camera.py
102 lines (86 loc) · 3.46 KB
/
recognition_camera.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
"""
desc: 利用摄像头实时检测
"""
import os
import argparse
import cv2
import numpy as np
from model import CNN
from utils import index2emotion, cv2_img_add_text
os.environ["TF_CPP_MIN_LOG_LEVEL"] = "2"
parser = argparse.ArgumentParser()
parser.add_argument("--source", type=int, default=0, help="data source, 0 for camera 1 for video")
parser.add_argument("--video_path", type=str, default=None)
opt = parser.parse_args()
if opt.source == 1 and opt.video_path is not None:
filename = opt.video_path
else:
filename = None
def load_model():
"""
加载本地模型
:return:
"""
model = CNN()
model.load_weights('./models/cnn3_best_weights.h5')
return model
def generate_faces(face_img, img_size=48):
"""
将探测到的人脸进行增广
:param face_img: 灰度化的单个人脸图
:param img_size: 目标图片大小
:return:
"""
face_img = face_img / 255.
face_img = cv2.resize(face_img, (img_size, img_size), interpolation=cv2.INTER_LINEAR)
resized_images = list()
resized_images.append(face_img)
resized_images.append(face_img[2:45, :])
resized_images.append(face_img[1:47, :])
resized_images.append(cv2.flip(face_img[:, :], 1))
for i in range(len(resized_images)):
resized_images[i] = cv2.resize(resized_images[i], (img_size, img_size))
resized_images[i] = np.expand_dims(resized_images[i], axis=-1)
resized_images = np.array(resized_images)
return resized_images
def predict_expression():
"""
实时预测
:return:
"""
# 参数设置
model = load_model()
border_color = (0, 0, 0) # 黑框
font_color = (255, 255, 255) # 白字
capture = cv2.VideoCapture(0) # 指定0号摄像头
if filename:
capture = cv2.VideoCapture(filename)
while True:
_, frame = capture.read() # 读取一帧视频，返回是否到达视频结尾的布尔值和这一帧的图像
# frame = cv2.cvtColor(cv2.resize(frame, (800, 600)), cv2.COLOR_BGR2RGB)
frame_gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY) # 灰度化
cascade = cv2.CascadeClassifier('./dataset/params/haarcascade_frontalface_alt.xml') # 检测人脸
# 利用分类器识别出哪个区域为人脸
faces = cascade.detectMultiScale(frame_gray, scaleFactor=1.1, minNeighbors=1, minSize=(120, 120))
# 如果检测到人脸
if len(faces) > 0:
for (x, y, w, h) in faces:
face = frame_gray[y: y + h, x: x + w] # 脸部图片
faces = generate_faces(face)
results = model.predict(faces)
result_sum = np.sum(results, axis=0).reshape(-1)
label_index = np.argmax(result_sum, axis=0)
emotion = index2emotion(label_index)
cv2.rectangle(frame, (x - 10, y - 10), (x + w + 10, y + h + 10), border_color, thickness=2)
frame = cv2_img_add_text(frame, emotion, x+30, y+30, font_color, 20)
# puttext中文显示问题
# cv2.putText(frame, emotion, (x + 30, y + 30), cv2.FONT_HERSHEY_SIMPLEX, 1, font_color, 4)
cv2.imshow("expression recognition(press esc to exit)", frame) # 利用人眼假象
key = cv2.waitKey(30) # 等待30ms，返回ASCII码
# 如果输入esc则退出循环
if key == 27:
break
capture.release() # 释放摄像头
cv2.destroyAllWindows() # 销毁窗口
if __name__ == '__main__':
predict_expression()