transforms + PIL实践裁剪图片和image和tensor的转换

截取了pix2pix源代码的aligned_dataset.py文件中的部分代码，进行单个图片测试

import random
import torchvision.transforms as transforms
from PIL import Image


AB = Image.open('fn18.jpg').convert('RGB') #利用Image大家打开图片，原图是一张宽高比为2：1的图片
print(AB.size)
#AB = AB.resize((512，256), Image.BICUBIC)
AB = transforms.ToTensor()(AB)#图片转成tensor
print(AB.shape)
w_total = AB.size(2)
w = int(w_total / 2)
h = AB.size(1)
print("w_total:%d, w:%d, h:%d"%(w_total,w,h))# w_total:512, w:256, h:256
#w_offset = random.randint(0, max(0, 29))#设置随机裁剪
#h_offset = random.randint(0, max(0, 29))
#A = AB[:, h_offset:h_offset + self.opt.fineSize,w_offset:w_offset + self.opt.fineSize]
#B = AB[:, h_offset:h_offset + self.opt.fineSize,w + w_offset:w + w_offset +self.opt.fineSize]
A = AB[:, 0:h, 0:w]#截取照片三个通道、从0~h的高，从0~w的宽
print(A.shape)
B = AB[:, 0:h, w:w_total]#截取照片三个通道、从0~h的高，从w~w_total的宽
print(B.shape)
#tensor转换成图片
imgAB = transforms.ToPILImage()(AB.float())
imgAB.show()
imgA = transforms.ToPILImage()(A.float())
imgA.show()
imgB = transforms.ToPILImage()(B.float())
imgB.show()
# A = transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))(A)#训练时进行归一化处理
# B = transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))(B)