from keras.layers import Input, MaxPool2D, Deconvolution2D, Convolution2D, Add, Dense, AveragePooling2D, UpSampling2D, Reshape, Flatten, Subtract, Concatenate
from keras.optimizers import Adam
from keras.callbacks import LearningRateScheduler
from keras import backend as k
from keras.utils import multi_gpu_model
import tensorflow as tf
from keras.utils import plot_model
from subpixel import Subpixel
from PIL import Image
import cv2
folder = '../Data'
CHANNEL = 3
DATA = DATA(folder=folder)
out_patch_size = DATA.patch_size
inp_patch_size = int(out_patch_size / 2)
DATA.load_data(folder=folder)
print("Training data Y:", DATA.training_patches_Y.shape)
print("Training data X:", DATA.training_patches_2x.shape)
print(" Max data: ", DATA.training_patches_2x.max())
def PSNRLossnp(y_true, y_pred):
return 10 * np.log(255 * 2 / (np.mean(np.square(y_pred - y_true))))
def SSIM(y_true, y_pred):
from keras.layers import Input, MaxPool2D, Deconvolution2D, Convolution2D, Add, Dense, AveragePooling2D, UpSampling2D, Reshape, Flatten, Subtract, Concatenate, Lambda
from keras.optimizers import Adam
from keras.callbacks import LearningRateScheduler
from keras import backend as k
from keras.applications import vgg16
from keras.utils import multi_gpu_model
from keras.applications.imagenet_utils import preprocess_input
import tensorflow as tf
from keras.utils import plot_model
from subpixel import Subpixel
from PIL import Image
import cv2
CHANNEL = 3
DATA = DATA(folder=folder, patch_size=int(scale * 32))
out_patch_size = DATA.patch_size
inp_patch_size = int(out_patch_size / scale)
if not test_only:
DATA.load_data(folder=folder)
if scale == 2:
x = DATA.training_patches_2x
elif scale == 4:
x = DATA.training_patches_4x
elif scale == 8:
x = DATA.training_patches_8x
def PSNRLossnp(y_true, y_pred):
return 10 * np.log(255 * 2 / (np.mean(np.square(y_pred - y_true))))
from keras.optimizers import Adam
from keras.callbacks import LearningRateScheduler
from keras import backend as k
from keras.utils import multi_gpu_model
import tensorflow as tf
from keras.utils import plot_model
from subpixel import Subpixel
from PIL import Image
import cv2
folder = '../BSDS200 processed'
CHANNEL = 3
DATA = DATA(folder = folder)
out_patch_size = DATA.patch_size
inp_patch_size = int(out_patch_size/ 2)
DATA.load_data(folder=folder)
print("Training data Y:" , DATA.training_patches_Y.shape)
print("Training data X:" , DATA.training_patches_2x.shape)
def PSNRLossnp(y_true,y_pred):
return 10* np.log(255*2 / np.mean(np.square(y_pred - y_true)))
def SSIM( y_true,y_pred):
u_true = k.mean(y_true)
u_pred = k.mean(y_pred)
# if j > epochs:
# break
def predict(self, p):
p = p / 255
g = self.get_model().predict(np.array(p))
g[0] = np.array(g[0] * 255, np.uint8)
g[1] = np.array(g[1] * 255, np.uint8)
g[2] = np.array(g[2] * 255, np.uint8)
return g
if __name__ == '__main__':
CHANNEL = 3
out_patch_size = 128
inp_patch_size = 16
DATA = DATA(folder=folder, patch_size=out_patch_size)
if not test_only:
DATA.load_data(folder=folder)
x = DATA.training_patches_8x
x2 = DATA.training_patches_4x
x4 = DATA.training_patches_2x
x8 = DATA.training_patches_Y
net = SRResnet(lr=learning_rate,
chk=chk,
test=test_only,
content_lambda=content_lambda,
mse_lambda=mse_lambda,
adv_lambda=adv_lambda,
patch_size=inp_patch_size,
visualize=not test_only)