Se usó la API de Cloud Translation para traducir esta página.
Switch to English

Transferencia rápida de estilos para estilos arbitrarios

Ver en TensorFlow.org Ejecutar en Google Colab Ver en GitHub Descargar cuaderno Ver modelo TF Hub

Basado en el código del modelo en magenta y la publicación:

Explorando la estructura de una red de estilización artística neuronal arbitraria en tiempo real . Golnaz Ghiasi, Honglak Lee, Manjunath Kudlur, Vincent Dumoulin, Jonathon Shlens , Actas de la Conferencia Británica de Visión Artificial (BMVC), 2017.

Preparar

Comencemos con la importación de TF-2 y todas las dependencias relevantes.

import functools
import os

from matplotlib import gridspec
import matplotlib.pylab as plt
import numpy as np
import tensorflow as tf
import tensorflow_hub as hub

print("TF Version: ", tf.__version__)
print("TF-Hub version: ", hub.__version__)
print("Eager mode enabled: ", tf.executing_eagerly())
print("GPU available: ", tf.test.is_gpu_available())
TF Version:  2.3.1
TF-Hub version:  0.9.0
Eager mode enabled:  True
WARNING:tensorflow:From <ipython-input-1-1b40332be70f>:13: is_gpu_available (from tensorflow.python.framework.test_util) is deprecated and will be removed in a future version.
Instructions for updating:
Use `tf.config.list_physical_devices('GPU')` instead.
GPU available:  True

# @title Define image loading and visualization functions  { display-mode: "form" }

def crop_center(image):
  """Returns a cropped square image."""
  shape = image.shape
  new_shape = min(shape[1], shape[2])
  offset_y = max(shape[1] - shape[2], 0) // 2
  offset_x = max(shape[2] - shape[1], 0) // 2
  image = tf.image.crop_to_bounding_box(
      image, offset_y, offset_x, new_shape, new_shape)
  return image

@functools.lru_cache(maxsize=None)
def load_image(image_url, image_size=(256, 256), preserve_aspect_ratio=True):
  """Loads and preprocesses images."""
  # Cache image file locally.
  image_path = tf.keras.utils.get_file(os.path.basename(image_url)[-128:], image_url)
  # Load and convert to float32 numpy array, add batch dimension, and normalize to range [0, 1].
  img = plt.imread(image_path).astype(np.float32)[np.newaxis, ...]
  if img.max() > 1.0:
    img = img / 255.
  if len(img.shape) == 3:
    img = tf.stack([img, img, img], axis=-1)
  img = crop_center(img)
  img = tf.image.resize(img, image_size, preserve_aspect_ratio=True)
  return img

def show_n(images, titles=('',)):
  n = len(images)
  image_sizes = [image.shape[1] for image in images]
  w = (image_sizes[0] * 6) // 320
  plt.figure(figsize=(w  * n, w))
  gs = gridspec.GridSpec(1, n, width_ratios=image_sizes)
  for i in range(n):
    plt.subplot(gs[i])
    plt.imshow(images[i][0], aspect='equal')
    plt.axis('off')
    plt.title(titles[i] if len(titles) > i else '')
  plt.show()

Consigamos también algunas imágenes para jugar.

# @title Load example images  { display-mode: "form" }

content_image_url = 'https://upload.wikimedia.org/wikipedia/commons/thumb/f/fd/Golden_Gate_Bridge_from_Battery_Spencer.jpg/640px-Golden_Gate_Bridge_from_Battery_Spencer.jpg'  # @param {type:"string"}
style_image_url = 'https://upload.wikimedia.org/wikipedia/commons/0/0a/The_Great_Wave_off_Kanagawa.jpg'  # @param {type:"string"}
output_image_size = 384  # @param {type:"integer"}

# The content image size can be arbitrary.
content_img_size = (output_image_size, output_image_size)
# The style prediction model was trained with image size 256 and it's the 
# recommended image size for the style image (though, other sizes work as 
# well but will lead to different results).
style_img_size = (256, 256)  # Recommended to keep it at 256.

content_image = load_image(content_image_url, content_img_size)
style_image = load_image(style_image_url, style_img_size)
style_image = tf.nn.avg_pool(style_image, ksize=[3,3], strides=[1,1], padding='SAME')
show_n([content_image, style_image], ['Content image', 'Style image'])
Downloading data from https://upload.wikimedia.org/wikipedia/commons/thumb/f/fd/Golden_Gate_Bridge_from_Battery_Spencer.jpg/640px-Golden_Gate_Bridge_from_Battery_Spencer.jpg
65536/58102 [=================================] - 0s 0us/step
Downloading data from https://upload.wikimedia.org/wikipedia/commons/0/0a/The_Great_Wave_off_Kanagawa.jpg
2686976/2684586 [==============================] - 0s 0us/step

png

Importar módulo TF-Hub

# Load TF-Hub module.

hub_handle = 'https://tfhub.dev/google/magenta/arbitrary-image-stylization-v1-256/2'
hub_module = hub.load(hub_handle)

La firma de este módulo central para la estilización de imágenes es:

outputs = hub_module(content_image, style_image)
stylized_image = outputs[0]

Donde se espera que content_image , style_image y stylized_image sean tensores 4-D con formas [batch_size, image_height, image_width, 3] .

En el ejemplo actual, solo proporcionamos imágenes individuales y, por lo tanto, la dimensión del lote es 1, pero se puede usar el mismo módulo para procesar más imágenes al mismo tiempo.

Los valores de entrada y salida de las imágenes deben estar en el rango [0, 1].

Las formas del contenido y la imagen de estilo no tienen por qué coincidir. La forma de la imagen de salida es la misma que la forma de la imagen del contenido.

Demuestre la estilización de la imagen

# Stylize content image with given style image.
# This is pretty fast within a few milliseconds on a GPU.

outputs = hub_module(tf.constant(content_image), tf.constant(style_image))
stylized_image = outputs[0]
# Visualize input images and the generated stylized image.

show_n([content_image, style_image, stylized_image], titles=['Original content image', 'Style image', 'Stylized image'])

png

Probémoslo en más imágenes

# @title To Run: Load more images { display-mode: "form" }

content_urls = dict(
  sea_turtle='https://upload.wikimedia.org/wikipedia/commons/d/d7/Green_Sea_Turtle_grazing_seagrass.jpg',
  tuebingen='https://upload.wikimedia.org/wikipedia/commons/0/00/Tuebingen_Neckarfront.jpg',
  grace_hopper='https://storage.googleapis.com/download.tensorflow.org/example_images/grace_hopper.jpg',
  )
style_urls = dict(
  kanagawa_great_wave='https://upload.wikimedia.org/wikipedia/commons/0/0a/The_Great_Wave_off_Kanagawa.jpg',
  kandinsky_composition_7='https://upload.wikimedia.org/wikipedia/commons/b/b4/Vassily_Kandinsky%2C_1913_-_Composition_7.jpg',
  hubble_pillars_of_creation='https://upload.wikimedia.org/wikipedia/commons/6/68/Pillars_of_creation_2014_HST_WFC3-UVIS_full-res_denoised.jpg',
  van_gogh_starry_night='https://upload.wikimedia.org/wikipedia/commons/thumb/e/ea/Van_Gogh_-_Starry_Night_-_Google_Art_Project.jpg/1024px-Van_Gogh_-_Starry_Night_-_Google_Art_Project.jpg',
  turner_nantes='https://upload.wikimedia.org/wikipedia/commons/b/b7/JMW_Turner_-_Nantes_from_the_Ile_Feydeau.jpg',
  munch_scream='https://upload.wikimedia.org/wikipedia/commons/c/c5/Edvard_Munch%2C_1893%2C_The_Scream%2C_oil%2C_tempera_and_pastel_on_cardboard%2C_91_x_73_cm%2C_National_Gallery_of_Norway.jpg',
  picasso_demoiselles_avignon='https://upload.wikimedia.org/wikipedia/en/4/4c/Les_Demoiselles_d%27Avignon.jpg',
  picasso_violin='https://upload.wikimedia.org/wikipedia/en/3/3c/Pablo_Picasso%2C_1911-12%2C_Violon_%28Violin%29%2C_oil_on_canvas%2C_Kr%C3%B6ller-M%C3%BCller_Museum%2C_Otterlo%2C_Netherlands.jpg',
  picasso_bottle_of_rum='https://upload.wikimedia.org/wikipedia/en/7/7f/Pablo_Picasso%2C_1911%2C_Still_Life_with_a_Bottle_of_Rum%2C_oil_on_canvas%2C_61.3_x_50.5_cm%2C_Metropolitan_Museum_of_Art%2C_New_York.jpg',
  fire='https://upload.wikimedia.org/wikipedia/commons/3/36/Large_bonfire.jpg',
  derkovits_woman_head='https://upload.wikimedia.org/wikipedia/commons/0/0d/Derkovits_Gyula_Woman_head_1922.jpg',
  amadeo_style_life='https://upload.wikimedia.org/wikipedia/commons/8/8e/Untitled_%28Still_life%29_%281913%29_-_Amadeo_Souza-Cardoso_%281887-1918%29_%2817385824283%29.jpg',
  derkovtis_talig='https://upload.wikimedia.org/wikipedia/commons/3/37/Derkovits_Gyula_Talig%C3%A1s_1920.jpg',
  amadeo_cardoso='https://upload.wikimedia.org/wikipedia/commons/7/7d/Amadeo_de_Souza-Cardoso%2C_1915_-_Landscape_with_black_figure.jpg'
)

content_image_size = 384
style_image_size = 256
content_images = {k: load_image(v, (content_image_size, content_image_size)) for k, v in content_urls.items()}
style_images = {k: load_image(v, (style_image_size, style_image_size)) for k, v in style_urls.items()}
style_images = {k: tf.nn.avg_pool(style_image, ksize=[3,3], strides=[1,1], padding='SAME') for k, style_image in style_images.items()}

Downloading data from https://upload.wikimedia.org/wikipedia/commons/d/d7/Green_Sea_Turtle_grazing_seagrass.jpg
3178496/3170828 [==============================] - 0s 0us/step
Downloading data from https://upload.wikimedia.org/wikipedia/commons/0/00/Tuebingen_Neckarfront.jpg
409600/406531 [==============================] - 0s 0us/step
Downloading data from https://storage.googleapis.com/download.tensorflow.org/example_images/grace_hopper.jpg
65536/61306 [================================] - 0s 0us/step
Downloading data from https://upload.wikimedia.org/wikipedia/commons/b/b4/Vassily_Kandinsky%2C_1913_-_Composition_7.jpg
196608/195196 [==============================] - 0s 0us/step
Downloading data from https://upload.wikimedia.org/wikipedia/commons/6/68/Pillars_of_creation_2014_HST_WFC3-UVIS_full-res_denoised.jpg
46931968/46930988 [==============================] - 2s 0us/step
Downloading data from https://upload.wikimedia.org/wikipedia/commons/thumb/e/ea/Van_Gogh_-_Starry_Night_-_Google_Art_Project.jpg/1024px-Van_Gogh_-_Starry_Night_-_Google_Art_Project.jpg
401408/396423 [==============================] - 0s 0us/step
Downloading data from https://upload.wikimedia.org/wikipedia/commons/b/b7/JMW_Turner_-_Nantes_from_the_Ile_Feydeau.jpg
147456/144340 [==============================] - 0s 0us/step
Downloading data from https://upload.wikimedia.org/wikipedia/commons/c/c5/Edvard_Munch%2C_1893%2C_The_Scream%2C_oil%2C_tempera_and_pastel_on_cardboard%2C_91_x_73_cm%2C_National_Gallery_of_Norway.jpg
11403264/11403121 [==============================] - 0s 0us/step
Downloading data from https://upload.wikimedia.org/wikipedia/en/4/4c/Les_Demoiselles_d%27Avignon.jpg
2908160/2905099 [==============================] - 0s 0us/step
Downloading data from https://upload.wikimedia.org/wikipedia/en/3/3c/Pablo_Picasso%2C_1911-12%2C_Violon_%28Violin%29%2C_oil_on_canvas%2C_Kr%C3%B6ller-M%C3%BCller_Museum%2C_Otterlo%2C_Netherlands.jpg
1236992/1234199 [==============================] - 0s 0us/step
Downloading data from https://upload.wikimedia.org/wikipedia/en/7/7f/Pablo_Picasso%2C_1911%2C_Still_Life_with_a_Bottle_of_Rum%2C_oil_on_canvas%2C_61.3_x_50.5_cm%2C_Metropolitan_Museum_of_Art%2C_New_York.jpg
122880/120288 [==============================] - 0s 0us/step
Downloading data from https://upload.wikimedia.org/wikipedia/commons/3/36/Large_bonfire.jpg
139264/131604 [===============================] - 0s 0us/step
Downloading data from https://upload.wikimedia.org/wikipedia/commons/0/0d/Derkovits_Gyula_Woman_head_1922.jpg
32768/32390 [==============================] - 0s 0us/step
Downloading data from https://upload.wikimedia.org/wikipedia/commons/8/8e/Untitled_%28Still_life%29_%281913%29_-_Amadeo_Souza-Cardoso_%281887-1918%29_%2817385824283%29.jpg
1916928/1914618 [==============================] - 0s 0us/step
Downloading data from https://upload.wikimedia.org/wikipedia/commons/3/37/Derkovits_Gyula_Talig%C3%A1s_1920.jpg
40960/40620 [==============================] - 0s 0us/step
Downloading data from https://upload.wikimedia.org/wikipedia/commons/7/7d/Amadeo_de_Souza-Cardoso%2C_1915_-_Landscape_with_black_figure.jpg
73728/66306 [=================================] - 0s 0us/step



content_name = 'sea_turtle'  # @param ['sea_turtle', 'tuebingen', 'grace_hopper']
style_name = 'munch_scream'  # @param ['kanagawa_great_wave', 'kandinsky_composition_7', 'hubble_pillars_of_creation', 'van_gogh_starry_night', 'turner_nantes', 'munch_scream', 'picasso_demoiselles_avignon', 'picasso_violin', 'picasso_bottle_of_rum', 'fire', 'derkovits_woman_head', 'amadeo_style_life', 'derkovtis_talig', 'amadeo_cardoso']

stylized_image = hub_module(tf.constant(content_images[content_name]),
                            tf.constant(style_images[style_name]))[0]

show_n([content_images[content_name], style_images[style_name], stylized_image],
       titles=['Original content image', 'Style image', 'Stylized image'])

png