-
Notifications
You must be signed in to change notification settings - Fork 0
/
img_colorization.py
30 lines (22 loc) · 1.03 KB
/
img_colorization.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
27
28
29
30
from processing import *
from decodingModel import *
# Using Transfer learning
feature_extract_model = VggModel()
#Decoding model
colorize = model()
#prepare data in hard disk
PrepareData(datapath = "/content/data" , save_file = "/content/processed/" ,
target_size = (224 , 224) , batch_size = 32 , feature_extract_model = feature_extract_model)
training_dir = "/content/processed"
num_train_samples = 1000
batch_size = 32
steps_per_epoch = np.floor(num_train_samples/batch_size)
epochs = 200
for i in range(epochs):
generator = data_generator_baseline(training_dir, num_train_samples, batch_size)
fit_history = colorize.fit_generator(generator, epochs=1, steps_per_epoch=steps_per_epoch, verbose=1)
if i % 10 == 0:
colorize.save('model_merge_' + str(i) + '.h5')
X = test_images(path = "/content/oldes" , shape = (224 , 224) , batch_size = 2 ,
feature_extract_model = feature_extract_model , model = colorize )
show_images(X , width = 20 , hight = 20 , columns = 2 , rows = 1)