针对图像分类的数据增强方法，离线增强，适合分类，无标签增强

针对图像分类的数据增强方法，离线增强，适合分类，无标签增强

代码：

改变路径即可使用

# 本代码主要提供一些针对图像分类的数据增强方法

# 1、平移。在图像平面上对图像以一定方式进行平移。
# 2、翻转图像。沿着水平或者垂直方向翻转图像。
# 3、旋转角度。随机旋转图像一定角度; 改变图像内容的朝向。
# 4、随机颜色。包括调整图像饱和度、亮度、对比度、锐度
# 5、缩放变形图片。
# 6、二值化图像。
# 7、随机黑色块遮挡
# 8、添加噪声

from PIL import Image
from PIL import ImageEnhance
from PIL import ImageChops
import os
import numpy as np


# 1、图像平移
def move(img): #平移，平移尺度为off
    offset = ImageChops.offset(img, np.random.randint(1, 20), np.random.randint(1, 40))
    return offset

# 2、翻转图像
def flip(img):   
    factor = np.random.randint(1, 3) #随机因子，随机上下或者左右翻转
    if factor == 1:
        filp_img = img.transpose(Image.FLIP_TOP_BOTTOM)
    else:
        filp_img = img.transpose(Image.FLIP_LEFT_RIGHT)
    return filp_img

#  3、旋转角度
def rotation(img):
    factor = np.random.randint(1, 21) #随机旋转角度
    rotation_img = img.rotate(factor) 
    return rotation_img

# 4、随机颜色 
def color(img): 
    """
    对图像进行颜色抖动
    :param image: PIL的图像image
    :return: 有颜色色差的图像image
    """
    random_factor = np.random.randint(5, 15) / 10.  # 随机因子
    color_image = ImageEnhance.Color(img).enhance(random_factor)                     # 调整图像的饱和度
    random_factor = np.random.randint(8, 15) / 10.  # 随机因子
    brightness_image = ImageEnhance.Brightness(color_image).enhance(random_factor)   # 调整图像的亮度
    random_factor = np.random.randint(10, 13) / 10. # 随机因子
    contrast_image = ImageEnhance.Contrast(brightness_image).enhance(random_factor)  # 调整图像对比度
    random_factor = np.random.randint(5, 31) / 10.  # 随机因子
    random_color = ImageEnhance.Sharpness(contrast_image).enhance(random_factor)     # 调整图像锐度
    return random_color 

# 5、缩放变形图片
def crop(img):
    factor_1 = np.random.randint(10, 50)
    factor_2 = np.random.randint(20, 50)
    crop_img = img.crop((img.size[0]/factor_1, img.size[1]/factor_2, img.size[0]*(factor_1-1)/factor_1, img.size[1]*(factor_2-1)/factor_2))
    cropResize_img = crop_img.resize((img.size[0], img.size[1]))
    return cropResize_img

# 6、二值化图像
def convert(img):
    convert_img = img.convert('L')
    return convert_img


# 7、黑色块遮挡
def paste(img):
    # 左上右下
    factor_1 = np.random.randint(20, 70)
    factor_2 = np.random.randint(30, 60)
    # 随机进行左边遮罩
    a = np.random.randint(1,3)
    if a == 2:
        img.paste((0,0,0),(int(img.size[0]*(factor_1-np.random.randint(2,4))/factor_1), 
                        int(img.size[1]*(np.random.randint(1,25))/factor_2), 
                        int(img.size[0]*(factor_1-np.random.randint(0,2))/factor_1),
                        int(img.size[1]*(np.random.randint(26,50))/factor_2)
                        ))
    else:
        # 随机进行底部遮罩
        img.paste((0,0,0),(int(img.size[0]*(np.random.randint(1,19))/factor_1), 
                        # int(img.size[1]*(factor_2-2)/factor_2), 
                        int(img.size[1]*(factor_2-np.random.randint(3,6))/factor_2),
                        int(img.size[0]*(np.random.randint(21,41))/factor_1),
                        # int(img.size[1]*(factor_2-1)/factor_2)
                        int(img.size[1]*(factor_2-np.random.randint(0,3))/factor_2)
                        ))
    return img


# 8、随机添加黑白噪声
def salt_and_pepper_noise(img, proportion = 0.00025):
    noise_img = img
    height,width =noise_img.size[0],noise_img.size[1]
    proportion = proportion * np.random.randint(1, 50)
    num = int(height * width * proportion) #多少个像素点添加椒盐噪声
    pixels = noise_img.load()
    for i in range(num):
        w = np.random.randint(0,width-1)
        h = np.random.randint(0,height-1)
        if np.random.randint(0,2) == 1:
            pixels[h,w] = 0
        else:
            pixels[h,w] = 255
    return noise_img



# 概率执行函数
def random_run(probability, func, useimage):
    """以probability%的概率执行func(*args)"""
    list = []
    for i in range(probability):
        list.append(1)                      #list中放入probability个1
    for x in range(100 - probability):
        list.append(0)                      #剩下的位置放入0
    a = np.random.choice(list)              #随机抽取一个
    if a == 0:
        return useimage
    if a == 1:
        image = func(useimage)
        return image




def main():
    imageDir = "D:/adavance/resnet50/datasets/Coupling/test/norm  "            #要改变的图片的路径文件夹
    saveDir = imageDir#"D:/adavance/resnet50/datasets/TailCotterPin/test/norm_TailCotterPin"                 #要保存的图片的路径文件夹
    seed = 10                           #每张初始图片要数据增强为多少张图片
    for name in os.listdir(imageDir):
        i=0
        for i in range(seed):
            i = i+1
            saveName = str(name[:-4]) + str(i) +".jpg"
            img = Image.open(os.path.join(imageDir, name))
            saveImage = random_run(60, flip, img)                               # 翻转
            saveImage = random_run(70, color, saveImage)                        # 色彩变化
            saveImage = random_run(30, crop, saveImage)                         # 裁减缩放
            #saveImage = random_run(30, paste, saveImage)                        # 添加遮罩
            saveImage = random_run(20, move, saveImage)                         # 平移
            saveImage = random_run(50, rotation, saveImage)                     # 旋转
            saveImage = random_run(10, convert, saveImage)                      # 二值化  
            saveImage = random_run(20, salt_and_pepper_noise, saveImage)        # 添加噪声点
            # saveImage = random_run(90, gauss_noise, saveImage)
            print(type(saveImage))
            if saveImage != None:
                saveImage.save(os.path.join(saveDir, saveName))
            else:
                pass
            print(i)


if __name__ == "__main__":
    main()


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

PS: 记得备份原图，要不然出差错了，就不好恢复了