# one convolution from 64 chanels to 3 chanels
generator = Conv2D(filters = 3, kernel_size = 9, strides = 1, padding = "same")(generator)
generator = Activation('sigmoid')(generator)
generator = Conv2D(filters = 32, kernel_size = 9, strides = 1, padding = "same")(generator)
generator = LeakyReLU(alpha = .2)(generator)
generator = Conv2D(filters = 64, kernel_size = 1, strides = 1, padding = "same")(generator)
generator = LeakyReLU(alpha = .2)(generator)
generator = Conv2D(filters = 3, kernel_size = 5, strides = 1, padding = "same")(generator)
generator = Activation("sigmoid")(generator)
# end of the graph creation and loading of the weights
generator = Model(inputs = input_layer, outputs = generator)
generator.load_weights("_generator.h5")
# load the image file and reshape color value between 0 and 1
image = Image.open(INPUT_FILE)
print(f"Image format = {image.getbands()}")
image = np.array(image).astype("float32") / 255
# extract the images data for th windows crations
x_size, y_size, _ = image.shape
new_size = x_size * 4, y_size * 4
# caluclate the windows number
x_size //= WINDOWS_SIZES
y_size //= WINDOWS_SIZES
x_size += 1
y_size += 1
discriminator = Dense(1)(discriminator)
discriminator = Activation('sigmoid')(discriminator)
# end of the graph creation
discriminator = Model(inputs=dis_input, outputs=discriminator)
# Compilation
#############
# generator
# the optimizer
optimizer = Adam(lr=2E-4, beta_1=0.9, beta_2=0.999, epsilon=1e-08)
# load & compilation
if LOAD:
generator.load_weights(SAVE_PATH + 'generator_model.h5')
_gnerator = generator # save an uncompile version of the generator
generator.compile(loss=vgg_loss, optimizer=optimizer)
# discriminator (use the same opti then the generator)
# load & compile
if LOAD:
discriminator.loss_weights(SAVE_PATH + 'discriminator_model.h5')
discriminator.compile(loss="binary_crossentropy",
optimizer=optimizer,
metrics=["accuracy"])
# gann
# the optimizer
