PIL Image，图像与tensor的转换，归一化

示例代码：

from torchvision import transforms
from PIL import Image

img_jpg = Image.open('C:/Users/admin/Desktop/bird.jpg').convert('RGB')   # convert()函数，用于不同模式图像之间的转换,
                                                                         # PIL中有九种不同模式，分别为1，L，P，RGB，RGBA，CMYK，YCbCr，I，F

print(img_jpg)   # 输出：<PIL.Image.Image image mode=RGB size=500x333 at 0x16ADDF8B948>

to_tensor = transforms.ToTensor()
img_tensor = to_tensor(img_jpg)    # img_tensor的每个通道最大值为1.0，最小值为0

normalize = transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))    # 归一化到[-1, 1]，公式是：(x-0.5)/0.5
img_norm = normalize(img_tensor)   # img_norm的每个通道最大值为1.0，最小值为-1.0


# transform tensor back to PIL image
img_unnorm = img_norm/2 + 0.5

to_PILimage = transforms.ToPILImage()
img_restored = to_PILimage(img_unnorm)

img_restored.save('C:/Users/admin/Desktop/bird_restored.jpg')

　　

图像：