PIL opencv 与tensor之间的相互转换

#--coding:utf-8--
import torch
from torchvision import transforms
from PIL import Image
import cv2
img_path = “./cat.59.jpg”
transform1 = transforms.Compose([
transforms.CenterCrop((224,224)), # 只能对PIL图片进行裁剪
transforms.ToTensor(),
]
)

PIL图片与Tensor互转

img_PIL = Image.open(img_path).convert(‘RGB’)
img_PIL.show() # 原始图片
img_PIL_Tensor = transform1(img_PIL)
print(type(img_PIL))
print(type(img_PIL_Tensor))

#Tensor转成PIL.Image重新显示
new_img_PIL = transforms.ToPILImage()(img_PIL_Tensor).convert(‘RGB’)
new_img_PIL.show() # 处理后的PIL图片

OPENCV图片与Tensor互转

#transforms中，没有对np.ndarray格式图像的操作
img_cv = cv2.imread(img_path)
transform2 = transforms.Compose([
transforms.ToTensor(),
]
)
img_cv_Tensor = transform2(img_cv)
print(type(img_cv))
print(type(img_cv_Tensor))