遥感影像语义分割——数据增强（图像和原图同时增强）

from PIL import Image, ImageFont, ImageDraw, ImageEnhance

import matplotlib.pyplot as plt

import numpy as np

import random

import  random

import os

def image_rotate(image,label):

    """

    对图像进行一定角度的旋转

    :param image_path:  图像路径

    :param save_path:   保存路径

    :param angle:       旋转角度

    :return:

    """

    image_rotated = image.transpose(Image.ROTATE_90).convert('RGB')

    label_rotated = label.transpose(Image.ROTATE_90)

    return image_rotated,label_rotated

def image_rotate1(image,label):

    """

    对图像进行一定角度的旋转

    :param image_path:  图像路径

    :param save_path:   保存路径

    :param angle:       旋转角度

    :return:

    """

    image_rotated = image.transpose(Image.ROTATE_270).convert('RGB')

    label_rotated = label.transpose(Image.ROTATE_270)

    return image_rotated,label_rotated

def bright(image):

    enh_bri = ImageEnhance.Brightness(image)

    brightness = 1.2

    image_brightened = enh_bri.enhance(brightness)

    return image_brightened.convert('RGB')

def ruidu(image):

    enh_sha = ImageEnhance.Sharpness(image)

    sharpness = 2.3

    image_sharped = enh_sha.enhance(sharpness)

    return image_sharped.convert('RGB')

def sedu(image):

    enh_col = ImageEnhance.Color(image)

    color = 1.2

    image_colored = enh_col.enhance(color)

    return image_colored.convert('RGB')

def duibidu(image):

    enh_con = ImageEnhance.Contrast(image)

    contrast = 1.3

    image_contrasted = enh_con.enhance(contrast)

    return image_contrasted.convert('RGB')

def image_flip(image,label):

    image_transpose = image.transpose(Image.FLIP_LEFT_RIGHT).convert('RGB')

    label_transpose = label.transpose(Image.FLIP_LEFT_RIGHT)

    return image_transpose,label_transpose

def image_color(image,label):

    image_transpose = image.transpose(Image.FLIP_TOP_BOTTOM).convert('RGB')

    label_transpose = label.transpose(Image.FLIP_TOP_BOTTOM)

    return image_transpose,label_transpose

path_img = r'E:\torch-deeplabv3\pytorch-deeplab-xception-master\Waste2021\JPEGImages'

path_label = r'E:\torch-deeplabv3\pytorch-deeplab-xception-master\Waste2021\SegmentationClass'

path_new_img = r'E:\aa\torch-deeplabv3\pytorch-deeplab-xception-master\Waste2021\JPEGImages'

path_new_label = r'E:\aa\torch-deeplabv3\pytorch-deeplab-xception-master\Waste2021\SegmentationClass'

img_list = os.listdir(path_img)

label_list = os.listdir(path_label)

k=0

for i in range(len(img_list)):

    img = Image.open(path_img + '/' + img_list[i])

    label =Image.open(path_label + '/' + img_list[i][0:-4] + '.png')

#保存原图

    img.convert('RGB').save(path_new_img + '/' + str(("%05d" % (k))) + '.jpg')

    label.save(path_new_label + '/' + str(("%05d" % (k))) + '.png')

    k += 1

#角度旋转第一次

    img1,mask = image_rotate(img,label)

    img1.save(path_new_img + '/' + str(("%05d" % (k))) + '.jpg')

    mask.save(path_new_label + '/' + str(("%05d" % (k))) + '.png')

    k+=1

#角度旋转第二次

    img1,mask = image_rotate1(img,label)

    img1.save(path_new_img + '/' + str(("%05d" % (k))) + '.jpg')

    mask.save(path_new_label + '/' + str(("%05d" % (k))) + '.png')

    k+=1

#调整亮度

    img1 = bright(img)

    img1.save(path_new_img + '/' + str(("%05d" % (k))) + '.jpg')

    label.save(path_new_label + '/' + str(("%05d" % (k))) + '.png')

    k+=1

#调整对比度

    img1 = duibidu(img)

    img1.save(path_new_img + '/' + str(("%05d" % (k))) + '.jpg')

    label.save(path_new_label + '/' + str(("%05d" % (k))) + '.png')

    k+=1

#调整锐度

    img1 = ruidu(img)

    img1.save(path_new_img + '/' + str(("%05d" % (k))) + '.jpg')

    label.save(path_new_label + '/' + str(("%05d" % (k))) + '.png')

    k+=1

#调整色度

    img1 = sedu(img)

    img1.save(path_new_img + '/' + str(("%05d" % (k))) + '.jpg')

    label.save(path_new_label + '/' + str(("%05d" % (k))) + '.png')

    k+=1

#左右翻转

    img1,mask = image_flip(img,label)

    img1.save(path_new_img + '/' + str(("%05d" % (k))) + '.jpg')

    mask.save(path_new_label + '/' + str(("%05d" % (k))) + '.png')

    k+=1

#上下翻转

    img1,mask = image_color(img,label)

    img1.save(path_new_img + '/' + str(("%05d" % (k))) + '.jpg')

    mask.save(path_new_label + '/' + str(("%05d" % (k))) + '.png')

    k += 1

    print(img_list[i] + 'is finished')