任意风格快速风格迁移

在 TensorFlow.org 上查看 在 Google Colab 中运行 在 GitHub 上查看 下载笔记本 查看 TF Hub 模型

基于 magenta 中的模型代码和出版物

探索实时任意神经艺术风格化网络的结构Golnaz Ghiasi、Honglak Lee、Manjunath Kudlur、Vincent Dumoulin、Jonathon Shlens,英国机器视觉会议 (BMVC) 论文集,2017 年。

设置

让我们从导入 TF2 和所有相关依赖项开始。

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.config.list_physical_devices('GPU'))
TF Version:  2.16.1
TF Hub version:  0.16.1
Eager mode enabled:  True
GPU available:  []
2024-03-10 11:57:42.713691: E external/local_xla/xla/stream_executor/cuda/cuda_driver.cc:282] failed call to cuInit: CUDA_ERROR_NO_DEVICE: no CUDA-capable device is detected
# @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 = tf.io.decode_image(
      tf.io.read_file(image_path),
      channels=3, dtype=tf.float32)[tf.newaxis, ...]
  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()

让我们也获取一些图像来玩玩。

# @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
71918/71918 ━━━━━━━━━━━━━━━━━━━━ 0s 0us/step
Downloading data from https://upload.wikimedia.org/wikipedia/commons/0/0a/The_Great_Wave_off_Kanagawa.jpg
2684586/2684586 ━━━━━━━━━━━━━━━━━━━━ 0s 0us/step

png

导入 TF Hub 模块

# Load TF Hub module.

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

此图像风格化 Hub 模块的签名为

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

其中 content_imagestyle_imagestylized_image 预计为形状为 [batch_size, image_height, image_width, 3] 的 4D 张量。

在当前示例中,我们只提供单个图像,因此批次维度为 1,但可以使用同一个模块同时处理更多图像。

图像的输入和输出值应在 [0, 1] 范围内。

内容图像和风格图像的形状不必匹配。输出图像形状与内容图像形状相同。

演示图像风格化

# 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

让我们在更多图像上尝试一下

# @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
3170828/3170828 ━━━━━━━━━━━━━━━━━━━━ 0s 0us/step
Downloading data from https://upload.wikimedia.org/wikipedia/commons/0/00/Tuebingen_Neckarfront.jpg
406531/406531 ━━━━━━━━━━━━━━━━━━━━ 0s 0us/step
Downloading data from https://storage.googleapis.com/download.tensorflow.org/example_images/grace_hopper.jpg
61306/61306 ━━━━━━━━━━━━━━━━━━━━ 0s 0us/step
Downloading data from https://upload.wikimedia.org/wikipedia/commons/b/b4/Vassily_Kandinsky%2C_1913_-_Composition_7.jpg
195196/195196 ━━━━━━━━━━━━━━━━━━━━ 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
46930988/46930988 ━━━━━━━━━━━━━━━━━━━━ 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
397382/397382 ━━━━━━━━━━━━━━━━━━━━ 0s 0us/step
Downloading data from https://upload.wikimedia.org/wikipedia/commons/b/b7/JMW_Turner_-_Nantes_from_the_Ile_Feydeau.jpg
144340/144340 ━━━━━━━━━━━━━━━━━━━━ 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
11403121/11403121 ━━━━━━━━━━━━━━━━━━━━ 0s 0us/step
Downloading data from https://upload.wikimedia.org/wikipedia/en/4/4c/Les_Demoiselles_d%27Avignon.jpg
2905099/2905099 ━━━━━━━━━━━━━━━━━━━━ 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
1234199/1234199 ━━━━━━━━━━━━━━━━━━━━ 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
120288/120288 ━━━━━━━━━━━━━━━━━━━━ 0s 0us/step
Downloading data from https://upload.wikimedia.org/wikipedia/commons/3/36/Large_bonfire.jpg
131604/131604 ━━━━━━━━━━━━━━━━━━━━ 0s 0us/step
Downloading data from https://upload.wikimedia.org/wikipedia/commons/0/0d/Derkovits_Gyula_Woman_head_1922.jpg
32390/32390 ━━━━━━━━━━━━━━━━━━━━ 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
1914618/1914618 ━━━━━━━━━━━━━━━━━━━━ 0s 0us/step
Downloading data from https://upload.wikimedia.org/wikipedia/commons/3/37/Derkovits_Gyula_Talig%C3%A1s_1920.jpg
40620/40620 ━━━━━━━━━━━━━━━━━━━━ 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
66306/66306 ━━━━━━━━━━━━━━━━━━━━ 0s 0us/step

指定要使用的主要内容图像和风格。

png