计算image1和image2之间的LPIPS指标的python代码_计算两个数据集的lpips

计算image1和image2之间的LPIPS指标的python代码

import cv2
import lpips
import torch

loss_fn_vgg = lpips.LPIPS(net='alex').to("cuda:0")

image1_path = 'F:\YXL\project\Restormer-mainV364_v1\YXL_dir\photo\photo_otput_result.png'
image2_path = 'F:\YXL\project\Restormer-mainV364_v1\YXL_dir\photo/target_decoder_otput_result.png'

img1 = cv2.imread(image1_path)
img2 = cv2.imread(image2_path)

restored = img2.transpose(2, 0, 1)
target = img1.transpose(2, 0, 1)

restored = torch.tensor(restored)
target = torch.tensor(target)

restored = restored.to("cuda:0")
target = target.to("cuda:0")

lpips_value = loss_fn_vgg(restored, target)

print("image1 and image2 ------- LPIPS: ", float(lpips_value))

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25