from implementations.support_scripts.common import h5_small_vgg_generator import numpy as np from skimage import io, color import matplotlib.pyplot as plt g = h5_small_vgg_generator(16, "../h5_data", None) b = next(g) vgg = b[0][1][4, :, :, :] s = b[0][0][4, :, :, :] c = b[1][4, :, :, :] im = np.concatenate((s, c), axis=2) rgb = color.lab2rgb(im) imgplot = plt.imshow(rgb) plt.show() plt.imshow(vgg[:, :, 0]) plt.show()
opt = optimizers.Adam(lr=1E-4, beta_1=0.9, beta_2=0.999, epsilon=1e-08)
model.compile(optimizer=opt, loss=custom_mse)
model.summary()
start_from = 75
save_every_n_epoch = 5
n_epochs = 10000
model.load_weights("../weights/implementation7d-relu-70.h5")
# start image downloader
# ip = ImagePacker("../small_dataset", "../h5_data", "imp7d-relu-", num_images=1024, num_files=None)
# ip.start()
ip = None
g = h5_small_vgg_generator(b_size, "../h5_data", ip)
gval = h5_small_vgg_generator(b_size, "../h5_validate", None)
for i in range(start_from // save_every_n_epoch,
n_epochs // save_every_n_epoch):
print("START", i * save_every_n_epoch, "/", n_epochs)
history = model.fit_generator(g,
steps_per_epoch=60000 / b_size,
epochs=save_every_n_epoch,
validation_data=gval,
validation_steps=(128 // b_size))
model.save_weights("../weights/implementation7d-relu-" +
str(i * save_every_n_epoch) + ".h5")
# save sample images
whole_image_check(model, 20,
model = Model(inputs=[main_input, vgg16.input], output=last)
opt = optimizers.Adam(lr=1E-4, beta_1=0.9, beta_2=0.999, epsilon=1e-08)
model.compile(optimizer=opt, loss=custom_mse, metrics=[root_mean_squared_error, mean_squared_error])
model.summary()
start_from = 0
save_every_n_epoch = 1
n_epochs = 10000
# model.load_weights("../weights/implementation9-bn-24.h5")
# start image downloader
ip = None
g = h5_small_vgg_generator(b_size, "../data/h5_small_train", ip)
gval = h5_small_vgg_generator(b_size, "../data/h5_small_validation", None)
for i in range(start_from // save_every_n_epoch, n_epochs // save_every_n_epoch):
print("START", i * save_every_n_epoch, "/", n_epochs)
history = model.fit_generator(g, steps_per_epoch=100000//b_size, epochs=save_every_n_epoch,
validation_data=gval, validation_steps=(10000//b_size))
model.save_weights("../weights/implementation9-bn-" + str(i * save_every_n_epoch) + ".h5")
# save sample images
whole_image_check_overlapping(model, 80, "imp9-bn-" + str(i * save_every_n_epoch) + "-")
# save history
output = open('../history/imp9-bn-{:0=4d}.pkl'.format(i * save_every_n_epoch), 'wb')
pickle.dump(history.history, output)