在pytorch的深度学习中，imgplt.imread()与cv2.imread()读取图片，并且如何使用class类和定义def函数

在pytorch1.11中进行图像的深度学习时，数据增强这块记录一下。主要通过“旋转”变换来学习class类的使用和def函数的写法。

import random
import scipy.ndimage as ndi
import cv2
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.image as imgplt


def transform_matrix_offset_center(matrix, x, y):
    o_x = float(x) / 2 + 0.5
    o_y = float(y) / 2 + 0.5
    offset_matrix = np.array([[1, 0, o_x], [0, 1, o_y], [0, 0, 1]])
    reset_matrix = np.array([[1, 0, -o_x], [0, 1, -o_y], [0, 0, 1]])
    transform_matrix = np.dot(np.dot(offset_matrix, matrix), reset_matrix)
    return transform_matrix


def apply_transform(x, 
					transform_matrix, 
					channel_index=2, 
					fill_mode='nearest', 
					cval=0., 
					order=1):
    x = np.rollaxis(x, channel_index, 0)
    final_affine_matrix = transform_matrix[:2, :2]
    final_offset = transform_matrix[:2, 2]
    channel_images = [ndi.interpolation.affine_transform(x_channel,
                                                         final_affine_matrix,
                                                         final_offset,
                                                         order=order,
                                                         mode=fill_mode,
                                                         cval=cval) for x_channel in x]
    x = np.stack(channel_images, axis=0)
    x = np.rollaxis(x, 0, channel_index + 1)
    return x


'''旋转变换'''
def rotation(x, 
			rg=20, 
			is_random=False, 
			row_index=0, 
			col_index=1, 
			channel_index=2, 
			fill_mode='nearest', 
			cval=0.,
            order=1):
    if is_random:
        theta = np.pi / 180 * np.random.uniform(-rg, rg)
    else:
        theta = np.pi / 180 * rg
    rotation_matrix = np.array(
        [[np.cos(theta), -np.sin(theta), 0], 
        [np.sin(theta), np.cos(theta), 0], 
        [0, 0, 1]])

    h, w = x.shape[row_index], x.shape[col_index]
    
    # 调用函数transform_matrix_offset_center()
    transform_matrix = transform_matrix_offset_center(rotation_matrix, h, w)
    
	# 调用函数apply_transform()
    x = apply_transform(
        x,
        transform_matrix,
        channel_index,
        fill_mode,
        cval,
        order)
    return x

'''数据增强'''
def image_augmentation(img, 
					   row_index=0, 
					   col_index=1, 
					   channel_index=2, 
					   is_rotate=False, 
					   is_crop=False,
                       is_flip=False, 
                       is_shift=False):
    row = img.shape[row_index]
    col = img.shape[col_index]
    channel = img.shape[channel_index]
    if is_rotate:
        rint = random.randint(0, 1)
        if rint == 0:
            img = rotation(img, 
            			   rg=10, 
            			   is_random=True, 
            			   row_index=row_index, 
            			   col_index=col_index,
                           channel_index=channel_index)
    
    return img


def fun(img):
    image = image_augmentation(img, is_rotate=True)
    # plt.imshow(image)
    # plt.show()
    # print(image.shape)
    return image


# 读取原始图像
img_0 = imgplt.imread('C:\\Users\\Administrator\\Desktop\\梅花.jpg', 1)
print('img_0.shape:', '\n', img_0.shape)
plt.imshow(img_0)
plt.show()


# 旋转
img_1 = fun(img=img_0)
print('img_1.shape:', '\n', img_1.shape)
plt.imshow(img_1)
plt.show()


# resize
img_2 = cv2.resize(img_0, (668, 668), (0.5, 0.25))
print('img_2.shape:', '\n', img_2.shape)
plt.imshow(img_2)
plt.show()


# 改变通道数
img_3 = np.transpose(img_0, (2, 0, 1))
print('img_3.shape:', '\n', img_3.shape)
# plt.imshow(img_3)
plt.show()

程序运行后的结果：

总结

1 图像变换的基本操作

在运行过程中，如果用cv2.imread()读取图片、且通过plt.imshow()、plt.show()来显示图片，图片显示有误，如下图：

import cv2
import matplotlib.pyplot as plt

# 读取原始图像
img_0 = cv2.imread('C:\\Users\\Administrator\\Desktop\\1.jpg', 1)
print('img_0.shape:', '\n', img_0.shape)
plt.imshow(img_0)
plt.show()

解决方法：
参考链接：matplotlib读取图片并显示
引入import matplotlib.image as imgplt，并用imgplt.imread读取图片，则显示正常，如下图：

import matplotlib.pyplot as plt
import matplotlib.image as imgplt

# 读取原始图像
img_0 = imgplt.imread('C:\\Users\\Administrator\\Desktop\\1.jpg', 1)
print('img_0.shape:', '\n', img_0.shape)
plt.imshow(img_0)
plt.show()

2 通过“旋转”变换来学习class类的使用和def函数的写法

import random
import scipy.ndimage as ndi
import cv2
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.image as imgplt


def transform_matrix_offset_center(matrix, x, y):
    o_x = float(x) / 2 + 0.5
    o_y = float(y) / 2 + 0.5
    offset_matrix = np.array([[1, 0, o_x], [0, 1, o_y], [0, 0, 1]])
    reset_matrix = np.array([[1, 0, -o_x], [0, 1, -o_y], [0, 0, 1]])
    transform_matrix = np.dot(np.dot(offset_matrix, matrix), reset_matrix)
    return transform_matrix


def apply_transform(x,
                    transform_matrix,
                    channel_index=2,
                    fill_mode='nearest',
                    cval=0.,
                    order=1):
    x = np.rollaxis(x, channel_index, 0)
    final_affine_matrix = transform_matrix[:2, :2]
    final_offset = transform_matrix[:2, 2]
    channel_images = [ndi.interpolation.affine_transform(x_channel,
                                                         final_affine_matrix,
                                                         final_offset,
                                                         order=order,
                                                         mode=fill_mode,
                                                         cval=cval) for x_channel in x]
    x = np.stack(channel_images, axis=0)
    x = np.rollaxis(x, 0, channel_index + 1)
    return x


def rotation(x,
             rg=20,
             is_random=False,
             row_index=0,
             col_index=1,
             channel_index=2,
             fill_mode='nearest',
             cval=0.,
             order=1):
    if is_random:
        theta = np.pi / 180 * np.random.uniform(-rg, rg)
    else:
        theta = np.pi / 180 * rg
    rotation_matrix = np.array(
                                [[np.cos(theta), -np.sin(theta), 0],
                                [np.sin(theta), np.cos(theta), 0],
                                [0, 0, 1]])
    h, w = x.shape[row_index], x.shape[col_index]
    transform_matrix = transform_matrix_offset_center(rotation_matrix, h, w)
    x = apply_transform(
        x,
        transform_matrix,
        channel_index,
        fill_mode,
        cval,
        order)
    return x

''' 定义类'''
class MyReader:
    def __init__(self, iscolor=True):
        self.iscolor = iscolor

    def img_augmentation(self, img, row_index=0, col_index=1, channel_index=2, is_rotate=False):
        if is_rotate:
            rint = random.randint(0, 1)
            if rint == 0:
                img = rotation(img,
                               rg=10,
                               is_random=True,
                               row_index=row_index,
                               col_index=col_index,
                               channel_index=channel_index)
        return img

    def train_reader(self, img):
        self.img = img
        image = cv2.cvtColor(self.img, cv2.IMREAD_COLOR)
        image = self.img_augmentation(image, is_rotate=True)
        plt.imshow(image)
        plt.show()
        print(image.shape)
        return image


iscolor = True
myreader = MyReader(iscolor=iscolor)  # 初始化类
img_0 = imgplt.imread('C:\\Users\\Administrator\\Desktop\\梅花.jpg')  
myreader.train_reader(img=img_0)  # 调用类中的函数

运行完后，图片显示如下：

3.注意：如果用cv2.imread()读取文件时，文件及文件名最好是英文，否则会报错！！！

img_0 = cv2.imread('C:\\Users\\Administrator\\Desktop\\梅花.jpg')

改为英文状态后，除了颜色与原图不一样外，还是基本能显示出图像的。

img_0 = cv2.imread('C:\\Users\\Administrator\\Desktop\\meihua.jpg')