def generate_image(im_width,im_height,norder,pickle1,set,random_string, state):
#print('\\\\')
trainer = DataTrainer()
stepper = Stepper()
img_sequencer = ImageSequencer()
parser = argparse.ArgumentParser(description='Collect and curate a dataset of images.')
parser.add_argument('directory', nargs=1)
parser.add_argument('--size', nargs=2, type=int)
parser.add_argument('--norder', nargs=1, type=int)
parser.add_argument('--pickle', nargs=1, type=bool)
#args = parser.parse_args()
#print('///')
#b= os.path.dirname('../templates/index.html')
#print(os.path.dirname(b))
directory = './data/training/images/raw/' + str(set) + '/'
im_width = im_width
im_height = im_height
norder = norder #args.norder[0]
randomGenerator = state
try:
should_pickle = pickle1
except:
should_pickle = False
pickle_file_name = "{width}-{height}-{norder}.pickle".format(width=im_width, height=im_height, norder=norder)
pickled_data = {}
preexisting_pickle = False
if should_pickle:
has_pickle = os.path.isfile(directory + pickle_file_name)
if has_pickle:
print('Loading Pickled Data...')
with open(directory + pickle_file_name, 'rb') as pickle_file:
pickled_data = pickle.load(pickle_file)
if pickled_data:
trained_data = pickled_data
preexisting_pickle = True
else:
image_set = []
for fn in os.listdir(directory):
if fn[0] != '.' and fn[-7:] !=".pickle":
image_set.append(directory + fn)
concat_text = ""
for image in image_set:
image = Image.open(image)
image_as_text = img_sequencer.sequence_image_to_text(image)
concat_text = concat_text + ' ' + image_as_text
concat_text.strip()
trained_data = trainer.train_text_data(
raw_text = concat_text,
order = norder,
)
# saves pickeled data
if should_pickle and preexisting_pickle == False:
print('Pickling data for later use...')
with open(directory + pickle_file_name, 'wb+') as pickle_file:
pickle.dump(trained_data, pickle_file)
pickle_file.close()
#print('Stepping Image Sequence...')
gen_seq = stepper.new_set_length_sequence(
model = trained_data,
steps = im_width * im_height,
state = randomGenerator
)
#print('Generating Image...')
image = img_sequencer.convert_text_to_image(gen_seq, im_width, im_height)
print('saving')
image.save("data/processed/images/" + str(set) + "-" + random_string + ".png")
image = Image.open(image)
image_as_text = img_sequencer.sequence_image_to_text(image)
concat_text = concat_text + ' ' + image_as_text
concat_text.strip()
trained_data = trainer.train_text_data(
raw_text = concat_text,
order = norder,
)
# saves pickeled data
if should_pickle and preexisting_pickle == False:
print('Pickling data for later use...')
with open(directory + pickle_file_name, 'wb+') as pickle_file:
pickle.dump(trained_data, pickle_file)
pickle_file.close()
print('Stepping Image Sequence...')
gen_seq = stepper.new_set_length_sequence(
model = trained_data,
steps = im_width * im_height
)
print('Generating Image...')
image = img_sequencer.convert_text_to_image(gen_seq, im_width, im_height)
image.save("data/processed/images/sample.png")
# print(gen_seq)
from Trainer import DataTrainer, ImageSequencer
from Stepper import Stepper
trainer = DataTrainer()
stepper = Stepper()
img_sequencer = ImageSequencer()
bauhaus_images = [
'bauhaus1.png', 'bauhaus2.png', 'bauhaus3.png', 'bauhaus4.png'
]
concat_text = ""
for image in bauhaus_images:
image = Image.open(
"data/training/images/cropped_and_reduced/{image}".format(image=image))
image_as_text = img_sequencer.sequence_image_to_text(image)
concat_text = concat_text + ' ' + image_as_text
concat_text.strip()
trained_data = trainer.train_text_data(
raw_text=concat_text,
order=100,
)
gen_seq = stepper.new_set_length_sequence(model=trained_data, steps=200 * 200)
image = img_sequencer.convert_text_to_image(gen_seq, 200, 200)
image.save("data/processed/images/bauhaus5.png")
# print(gen_seq)