-
Notifications
You must be signed in to change notification settings - Fork 0
/
evaluation.py
26 lines (25 loc) · 1.09 KB
/
evaluation.py
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
26
import numpy as np
from tensorflow.keras.preprocessing.image import img_to_array, load_img
from skimage.transform import resize
from skimage.color import rgb2lab, gray2rgb
def test_model(test_path, file, transfer_learning_model, model):
test_org = img_to_array(load_img(test_path+file))
test = resize(test_org, (224, 224), anti_aliasing=True)
test_tl = resize(test_org, (299, 299), anti_aliasing=True)
test *= 1.0/255
test_tl *= 1.0/255
lab_test = rgb2lab(test)
lab_test_tl = rgb2lab(test_tl)
luminance_test = lab_test[:, :, 0]
luminance_test_tl = lab_test_tl[:, :, 0]
luminance_test_4d = luminance_test.reshape((1, 224, 224, 1))
embedings_test = gray2rgb(luminance_test_tl)
embedings_test = embedings_test.reshape((1, 299, 299, 3))
embedings_test = transfer_learning_model.predict(embedings_test)
ab_test = model.predict([embedings_test, luminance_test_4d])
ab_test = ab_test*128
cur = np.zeros((224, 224, 3))
cur[:, :, 0] = luminance_test
cur[:, :, 1:] = ab_test
cur = resize(cur, (512, 512))
return test_org, luminance_test, cur