数据处理之数据扩充图片数据增强

数据扩充图片数据增强

在训练深度学习模型中经常会出现数据量小导致模型训练效果不好的种种原因，其中有两种方法加以改进一种是对数据进行扩充然后在放入模型进行训练，即给数据添加噪声，亮度，暗度对比度等（感觉旋转效果比较一般），一种是对数据训练时候调用pytorch库对数据进行优化.我们这里使用第一种.
文件夹目录如下：

# -*- coding: utf-8 -*-

import cv2
import os
import numpy as np

'''
    1. 水平垂直翻转
    2. 高斯噪声
    3. 45°旋转
    4. 昏暗
'''
# 椒盐噪声
def SaltAndPepper(src,percetage):
    SP_NoiseImg=src.copy()
    SP_NoiseNum=int(percetage*src.shape[0]*src.shape[1])
    for i in range(SP_NoiseNum):
        randR=np.random.randint(0,src.shape[0]-1)
        randG=np.random.randint(0,src.shape[1]-1)
        randB=np.random.randint(0,3)
        if np.random.randint(0,1)==0:
            SP_NoiseImg[randR,randG,randB]=0
        else:
            SP_NoiseImg[randR,randG,randB]=255
    return SP_NoiseImg



# 添加高斯噪声
def gaussian_noise(img, mean=0, sigma=0.1):
    '''
    此函数用将产生的高斯噪声加到图片上
    均值为0，是保证图像的亮度不会有变化，而方差大小则决定了高斯噪声的强度。
    方差/标准差越大，噪声越强。

    传入参数:
        img   :  原图
        mean  :  均值
        sigma :  标准差
    返回值:
        gaussian_out : 噪声处理后的图片
    '''
    # 将图片灰度标准化
    img = img / 255
    # 产生高斯 noise
    noise = np.random.normal(mean, sigma, img.shape)
    # 将噪声和图片叠加
    gaussian_out = img + noise
    # 将超过 1 的置 1，低于 0 的置 0
    gaussian_out = np.clip(gaussian_out, 0, 1)
    # 将图片灰度范围的恢复为 0-255
    gaussian_out = np.uint8(gaussian_out * 255)
    # 将噪声范围搞为 0-255
    # noise = np.uint8(noise*255)
    return gaussian_out  # 这里也会返回噪声，注意返回值


# 变暗
def darker(image, percetage=0.9):
    image_copy = image.copy()
    w = image.shape[1]
    h = image.shape[0]
    # get darker
    for xi in range(0, w):
        for xj in range(0, h):
            image_copy[xj, xi, 0] = int(image[xj, xi, 0] * percetage)
            image_copy[xj, xi, 1] = int(image[xj, xi, 1] * percetage)
            image_copy[xj, xi, 2] = int(image[xj, xi, 2] * percetage)
    return image_copy


# 变亮
def brighter(image, percetage=1.5):
    image_copy = image.copy()
    w = image.shape[1]
    h = image.shape[0]
    # get brighter
    for xi in range(0, w):
        for xj in range(0, h):
            image_copy[xj, xi, 0] = np.clip(int(image[xj, xi, 0] * percetage), a_max=255, a_min=0)
            image_copy[xj, xi, 1] = np.clip(int(image[xj, xi, 1] * percetage), a_max=255, a_min=0)
            image_copy[xj, xi, 2] = np.clip(int(image[xj, xi, 2] * percetage), a_max=255, a_min=0)
    return image_copy


# 水平垂直翻转
def flip(pic):
    '''
    1 : 水平翻转
    0 ； 垂直翻转
    -1： 水平垂直翻转
    '''
    img = cv2.flip(pic, -1)
    return img


# 45度旋转
def rotate(pic):
    rows, cols = pic.shape[:2]
    M = cv2.getRotationMatrix2D((cols / 2, rows / 2), 45, 1)
    pic = cv2.warpAffine(pic, M, (cols, rows))
    return pic


# oldpath: 原始图像位置
# oldpath:处理后的图像位置
oldpath = "./silk"
newpath = "./enforceshi"
number = 0

# 遍历训练集的每张图片
folders = os.listdir(oldpath)
print("当前类：",folders)
for folder in folders:
    for image in os.listdir(os.path.join(oldpath, folder)):
        img = cv2.imread(os.path.join(oldpath, folder, image))
        print("ORIGINAL PICTURE:：",os.path.join(oldpath, folder, str(image[0:-4]) + '.jpg'))
        cv2.imwrite(os.path.join(newpath, folder, str(image[0:-4]) + '.jpg'),img)

        # 添加高斯噪声
        img_gaussnoise = gaussian_noise(img, mean=0, sigma=0.1)
        print("NEW PATH::",os.path.join(newpath, folder, "gaussnoise"+str(image[0:-4]) + '.jpg'))
        cv2.imwrite(os.path.join(newpath, folder, "gaussnoise"+str(image[0:-4]) + '.jpg'), img_gaussnoise)

        # 增加噪声
        img_salt = SaltAndPepper(img, 0.3)
        # cv2.imwrite(file_dir + img_name[0:7] + '_salt.jpg', img_salt)
        cv2.imwrite(os.path.join(newpath, folder, "salt" + str(image[0:-4]) + '.jpg'), img_salt)

        # # 水平垂直翻转
        # img_flip = flip(img)
        # cv2.imwrite(os.path.join(newpath, folder, str(image[0:-4]) + '_flip.jpg'), img_flip)
        #
        # # 旋转45°
        # img_rotate45 = rotate(img)
        # cv2.imwrite(os.path.join(newpath, folder, str(image[0:-4]) + '_rotate45.jpg'), img_rotate45)

        # 降低亮度
        img_darker = darker(img, percetage=0.8)
        cv2.imwrite(os.path.join(newpath, folder, "darker"+str(image[0:-4]) + '.jpg'), img_darker)

        # 提高亮度
        img_brighter = brighter(img, percetage=1.5)
        cv2.imwrite(os.path.join(newpath, folder, "brighter"+str(image[0:-4]) + '.jpg'), img_brighter)

        number += 1
        print('第{0}张图片处理完成'.format(number))

print("全部转换完成！")

效果如下：

参考：
https://blog.youkuaiyun.com/weixin_51391591/article/details/119607028
https://blog.youkuaiyun.com/qq_36756866/article/details/108255715