-
Notifications
You must be signed in to change notification settings - Fork 7
/
test.py
138 lines (130 loc) · 5.15 KB
/
test.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
import tensorflow as tf
import os
# 获取图片名字列表，图片流列表
def get_images():
# 图片名字列表
images_name = []
# 图片流列表
images_list = []
# 输入路径
# path = input("image path: ")
path = ".\\test_photos"
if os.path.isdir(path):
for dir in os.listdir(path):
for file in os.listdir(path + "\\" + dir):
if (os.path.isfile(path + "\\" + dir + "\\" + file)):
if os.path.splitext(file)[1] == ".jpg":
# 原始图片
image_raw = tf.io.read_file(path + "\\" + dir + "\\" + file)
# 原始图片转换为tensor类型
image_tensor = tf.image.decode_image(image_raw, channels=3)
# 修改尺寸
image_final = tf.image.resize(image_tensor, [192, 192])
# 绝对色彩信息
image_final = image_final / 255.0
# 升维
image_final = image_final[tf.newaxis, ...]
# 添加到图片名字列表
images_name.append(path + "\\" + dir + "\\" + file)
# 添加到图片流列表
images_list.append(image_final)
else:
# 原始图片
image_raw = tf.io.read_file(path)
# 原始图片转换为tensor类型
image_tensor = tf.image.decode_image(image_raw, channels=3)
# 修改尺寸
image_final = tf.image.resize(image_tensor, [192, 192])
# 绝对色彩信息
image_final = image_final / 255.0
# 升维
image_final = image_final[tf.newaxis, ...]
# 添加到图片名字列表
images_name.append(path)
# 添加到图片流列表
images_list.append(image_final)
# 获取图片名字列表，图片流列表
return images_name, images_list
# 根据文件夹名字获取标签
def get_label_name(path):
# 标签左开头
left = len(path) - 1
# 标签右结尾
right = len(path) - 1
# 是否已遍历过’/‘
flag = 1
# 从后往前遍历字符串
for i in range(len(path) - 1, -1, -1):
# 是否是‘/’
if (path[i] == '\\'):
# 未遍历过’/‘
if (flag):
# 标签右结尾
right = i
flag = 0
# 遍历过’/‘
else:
# 标签左开头
left = i + 1
break
# 返回标签名字
return path[left:right]
# 标签名字英文
labels_name = ['cherry', 'chinese rose', 'daisy', 'dandelion', 'myosotis', 'poppy', 'roses', 'sunflowers', 'tulips', 'violet']
# 标签名字中文
chinese_name = ['樱花', '月季', '雏菊', '蒲公英', '勿忘我', '罂粟花', '玫瑰', '向日葵', '郁金香', '紫罗兰']
# 模型存储位置
checkpoint_save_path = "./plant.ckpt"
# 获取已存在的MobileNetV2模型
mobile_net = tf.keras.applications.MobileNetV2(input_shape=(192, 192, 3), include_top=False)
# MobileNet的权重为不可训练
mobile_net.trainable = False
# 模型层次：mobile_net，平均池，全连接
model = tf.keras.Sequential([
mobile_net,
tf.keras.layers.GlobalAveragePooling2D(),
tf.keras.layers.Dense(len(labels_name), activation='softmax')])
# 判断是否已有模型
if os.path.exists(checkpoint_save_path + '.index'):
print('-------------加载模型-----------------')
# 加载现有模型
model.load_weights(checkpoint_save_path)
# 获取图片名字列表，图片流列表
images_name, images_stream = get_images()
# 各种类图片标签命中
image_acc = list(0 for i in range(len(labels_name)))
# 各种类图片计数
image_count = list(0 for i in range(len(labels_name)))
# 测试所有图片
for i in range(len(images_name)):
# 换行
print()
# 测试
result = model.predict(images_stream[i])
# 实际标签
real_label = get_label_name(images_name[i])
# 预测得到的标签
pre_label = labels_name[int(result.argmax())]
# 实际标签与预测得到的标签相同
if (real_label == pre_label):
image_acc[int(result.argmax())] = image_acc[int(result.argmax())] + 1
image_count[int(result.argmax())] = image_count[int(result.argmax())] + 1
# 实际标签与预测得到的标签不相同
else:
# 查找图片实际标签，并对图片计数加一
for i in range(len(labels_name)):
if (labels_name[i] == real_label):
image_count[i] = image_count[i] + 1
break
# 输出最大概率的标签
print(images_name[i] + "\n识别结果: " + pre_label)
# 输出标签所有概率
for j in range(len(labels_name)):
print(chinese_name[j] + ": " + str(round(result[0][j] * 100, 2)) + "%")
# 输出全部图片预测的准确率汇总
print("---------------所有图片统计信息---------------")
for i in range(len(labels_name)):
if (image_count[i]):
print(chinese_name[i] + "命中率: " + str(round(image_acc[i] / image_count[i] * 100, 2)) + "%")
else:
print(chinese_name[i] + "命中率: 0.00%")