-
Notifications
You must be signed in to change notification settings - Fork 1
/
merge.py
96 lines (83 loc) · 4.62 KB
/
merge.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
import cv2
import tensorflow as tf
# import serial
# from velocity_decision import velocity_process, serial_trans
from camera import WebcamVideoStream, configs
# from receiveData import data_acquire
import datetime
PATH_TO_CKPT = 'D:\Scene_Recognition_V6\pb_model\CNN-8.pb' # 训练好的网络模型参数文件路径
def main():
my_list = []
insert_count = 0
# function : 加载训练好模型，进行预测测试
load_start = datetime.datetime.now()
load_graph = tf.Graph()
with load_graph.as_default(): # 加载模型
od_graph_def = tf.GraphDef()
with tf.gfile.GFile(PATH_TO_CKPT, 'rb') as fid:
serialized_graph = fid.read()
od_graph_def.ParseFromString(serialized_graph)
tf.import_graph_def(od_graph_def, name='')
load_end = datetime.datetime.now()
print("加载模型用时：" + str((load_end - load_start).microseconds / 1000) + "毫秒")
while True:
# tf.reset_default_graph()
start = datetime.datetime.now()
frame_original = video_capture.read()
end1 = datetime.datetime.now()
# print("openCV读图用时" + str((end1 - start).microseconds / 1000) + "毫秒")
frame = cv2.resize(frame_original, (64, 64))
image = frame*(1./255) - 0.5
with load_graph.as_default():
with tf.Session(graph=load_graph) as sess:
example_test = image.reshape((1, 64, 64, 3)) # reshape 为 4D 张量
x = load_graph.get_tensor_by_name('images:0')
keep_prob1 = load_graph.get_tensor_by_name('Fc1/my_keep_prob:0')
keep_prob2 = load_graph.get_tensor_by_name('Fc2/my_keep_prob:0')
prediction = load_graph.get_tensor_by_name('Fc3/softmax/my_prediction:0')
end4 = datetime.datetime.now()
# pre = sess.run(prediction, feed_dict={x: example_test, keep_prob: 1.0})
pre = sess.run(prediction, feed_dict={x: example_test, keep_prob1: 1.0, keep_prob2: 1.0})
end5 = datetime.datetime.now()
print("sess.run(prediction): " + str((end5 - end4).microseconds / 1000) + "毫秒")
# prob_left = pre[0, 0]
# prob_right = pre[0, 1]
# prob_straight = pre[0, 2]
# print("prob_left：" + str(prob_left) + " prob_right：" + str(prob_right) + " prob_straight：" + str(prob_straight))
# frame = cv2.resize(frame_original, (520, 520))
#
# cv2.putText(frame, 'go_left: '+str(round(prob_left, 5)), (10, 40), 0, 1, (0, 0, 255), 2) # 显示预测的标签号在图片上
# cv2.putText(frame, 'go_right: ' + str(round(prob_right, 5)), (10, 80), 0, 1, (255, 0, 0), 2)
# cv2.putText(frame, 'go_straight: ' + str(round(prob_straight, 5)), (10, 120), 0, 1, (0, 255, 0), 2)
#
# line, angu = velocity_process(prob_left, prob_right, prob_straight)
# cv2.putText(frame, 'cmd V: ' + str(round(line, 3)) + ' m/s', (10, 470), 0, 1, (0, 0, 0), 2)
# cv2.putText(frame, 'cmd W: ' + str(round(angu, 3)) + ' rad/s', (10, 510), 0, 1, (0, 0, 0), 2)
#
# cv2.namedWindow("image", 1) # 定义显示 窗口
cv2.imshow("image", frame)
cv2.waitKey(1)
# serial_trans(ser, line, angu)
end = datetime.datetime.now()
classify_time = str((end - end1).microseconds / 1000)
print("模型分类用时" + classify_time + "毫秒")
frame_time = str((end - start).microseconds / 1000)
print("processing time: " + frame_time + "毫秒\r\n")
# classify_time = float(classify_time)
# frame_time = float(frame_time)
# val_temp = [classify_time, frame_time]
# my_list.append(val_temp) # 向数组中添加当前接收到的数据
# array_length = len(my_list) # 提取当前数组存储的数据量
# if array_length >= 500:
# insert_count = insert_count + 1
# save_time(my_list, insert_count)
# my_list = []
# print("分类时间已保存")
tf.get_default_graph().finalize()
if __name__ == '__main__':
args = configs()
video_capture = WebcamVideoStream(src=args.video_source).start()
# ser = serial.Serial("COM6", 115200, timeout=5)
# print("串口初始化完成！")
# data_acquire(ser)
main()