# model = models.ResNetSR(scale).create_model()
# plot_model(model, to_file="architectures/ResNet.png", show_layer_names=True, show_shapes=True)
# model = models.GANImageSuperResolutionModel(scale).create_model(mode='train')
# plot_model(model, to_file='architectures/GAN Image SR.png', show_shapes=True, show_layer_names=True)
# model = models.DistilledResNetSR(scale).create_model()
# plot_model(model, to_file='architectures/distilled_resnet_sr.png', show_layer_names=True, show_shapes=True)
# model = models.NonLocalResNetSR(scale).create_model()
# plot_model(model, to_file='architectures/non_local_resnet_sr.png', show_layer_names=True, show_shapes=True)
"""
Train Super Resolution
"""
sr = models.ImageSuperResolutionModel(scale)
sr.create_model()
sr.fit(nb_epochs=10, save_history=False) #kpl from 250
"""
Train ExpantionSuperResolution
"""
# esr = models.ExpantionSuperResolution(scale)
# esr.create_model()
# esr.fit(nb_epochs=250)
"""
Train DenoisingAutoEncoderSR
"""
# dsr = models.DenoisingAutoEncoderSR(scale)
# dsr.create_model()
from __future__ import print_function, division
from keras.utils.vis_utils import plot_model
import models
import img_utils
if __name__ == "__main__":
path = r"headline_carspeed.jpg"
val_path = "val_images/"
scale = 4
"""
Plot the models
"""
model = models.ImageSuperResolutionModel(scale).create_model()
plot_model(model,
to_file="architectures/SRCNN.png",
show_shapes=True,
show_layer_names=True)
# model = models.ExpantionSuperResolution(scale).create_model()
# plot_model(model, to_file="architectures/ESRCNN.png", show_layer_names=True, show_shapes=True)
# model = models.DenoisingAutoEncoderSR(scale).create_model()
# plot_model(model, to_file="architectures/Denoise.png", show_layer_names=True, show_shapes=True)
# model = models.DeepDenoiseSR(scale).create_model()
# plot_model(model, to_file="architectures/Deep Denoise.png", show_layer_names=True, show_shapes=True)
# model = models.ResNetSR(scale).create_model()
'"ddsr" or "rnsr"'
mode = str(args.mode).lower()
assert mode in ['fast',
'patch'], 'Mode of operation must be either "fast" or "patch"'
scale_factor = int(args.scale)
save = strToBool(args.save)
patch_size = int(args.patch_size)
assert patch_size > 0, "Patch size must be a positive integer"
customWeightsPath = args.weightsPath
if model_type == "sr":
model = models.ImageSuperResolutionModel(scale_factor, customWeightsPath)
elif model_type == "esr":
model = models.ExpantionSuperResolution(scale_factor)
elif model_type == "dsr":
model = models.DenoisingAutoEncoderSR(scale_factor)
elif model_type == "ddsr":
model = models.DeepDenoiseSR(scale_factor)
elif model_type == "rnsr":
model = models.ResNetSR(scale_factor)
else:
model = models.DeepDenoiseSR(scale_factor)
model.upscale(path,
save_intermediate=save,
mode=mode,
patch_size=patch_size,
model_type = str(args.model).lower()
assert model_type in ["sr", "esr", "dsr", "ddsr", "rnsr"], 'Model type must be either "sr", "esr", "dsr", ' \
'"ddsr" or "rnsr"'
mode = str(args.mode).lower()
assert mode in ['fast',
'patch'], 'Mode of operation must be either "fast" or "patch"'
scale_factor = int(args.scale)
save = strToBool(args.save)
patch_size = int(args.patch_size)
assert patch_size > 0, "Patch size must be a positive integer"
if model_type == "sr":
model = models.ImageSuperResolutionModel(scale_factor)
elif model_type == "esr":
model = models.ExpantionSuperResolution(scale_factor)
elif model_type == "dsr":
model = models.DenoisingAutoEncoderSR(scale_factor)
elif model_type == "ddsr":
model = models.DeepDenoiseSR(scale_factor)
elif model_type == "rnsr":
model = models.ResNetSR(scale_factor)
else:
model = models.DeepDenoiseSR(scale_factor)
model.upscale(path,
save_intermediate=save,
mode=mode,
patch_size=patch_size,
args = parser.parse_args()
path = args.imgpath
suffix = args.suffix
model_type = str(args.model).lower()
assert model_type in ["sr", "esr", "dsr", "ddsr", "rnsr"], 'Model type must be either "sr", "esr", "dsr", ' \
'"ddsr" or "rnsr"'
mode = str(args.mode).lower()
assert mode in ['fast', 'patch'], 'Mode of operation must be either "fast" or "patch"'
scale_factor = int(args.scale)
save = strToBool(args.save)
patch_size = int(args.patch_size)
assert patch_size > 0, "Patch size must be a positive integer"
if model_type == "sr":
model = models.ImageSuperResolutionModel()
elif model_type == "esr":
model = models.ExpantionSuperResolution()
elif model_type == "dsr":
model = models.DenoisingAutoEncoderSR()
elif model_type == "ddsr":
model = models.DeepDenoiseSR()
elif model_type == "rnsr":
model = models.ResNetSR()
model.upscale(path, scale_factor=scale_factor, save_intermediate=save, evaluate=False, mode=mode,
patch_size=patch_size, suffix=suffix)