Générer des images avec BigBiGAN

Voir sur TensorFlow.org Exécuter dans Google Colab Voir sur GitHub Télécharger le cahier Voir les modèles TF Hub

Ce bloc - notes est une démo pour les modèles BigBiGAN disponibles sur TF Hub .

BigBiGAN étend GAN standard (BIG) en ajoutant un module de codeur qui peut être utilisé pour l' apprentissage d' une représentation sans surveillance. Grosso modo, le codeur invertis le générateur en prédisant latentes z donné des données réelles x . Voir le papier BigBiGAN sur arXiv [1] pour plus d' informations sur ces modèles.

Après vous être connecté à un environnement d'exécution, commencez en suivant ces instructions :

  1. (Facultatif) Mise à jour sélectionné module_path dans la première cellule de code ci - dessous pour charger un générateur BigBiGAN pour une architecture de codeur différent.
  2. Cliquez sur Runtime> Exécuter tout pour exécuter chaque cellule dans l' ordre. Ensuite, les sorties, y compris les visualisations d'échantillons et de reconstructions BigBiGAN, devraient automatiquement apparaître ci-dessous.

[1] Jeff Donahue et Karen Simonyan. À grande échelle accusatoire apprentissage Représentation . arXiv: 1907,02544, 2019.

Tout d'abord, définissez le chemin du module. Par défaut, on charge le modèle BigBiGAN avec le plus petit codeur basé ResNet-50 de https://tfhub.dev/deepmind/bigbigan-resnet50/1 . Pour charger le modèle à base Revnet-50 x4 plus utilisé pour obtenir les meilleurs résultats d'apprentissage de la représentation, le commentaire actif module_path réglage et décommenter l'autre.

module_path = 'https://tfhub.dev/deepmind/bigbigan-resnet50/1'  # ResNet-50
# module_path = 'https://tfhub.dev/deepmind/bigbigan-revnet50x4/1'  # RevNet-50 x4

Installer

import io
import IPython.display
import PIL.Image
from pprint import pformat

import numpy as np

import tensorflow.compat.v1 as tf
tf.disable_v2_behavior()

import tensorflow_hub as hub
2021-07-29 11:33:37.869626: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudart.so.11.0
WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.7/site-packages/tensorflow/python/compat/v2_compat.py:96: disable_resource_variables (from tensorflow.python.ops.variable_scope) is deprecated and will be removed in a future version.
Instructions for updating:
non-resource variables are not supported in the long term

Définir quelques fonctions pour afficher des images

def imgrid(imarray, cols=4, pad=1, padval=255, row_major=True):
  """Lays out a [N, H, W, C] image array as a single image grid."""
  pad = int(pad)
  if pad < 0:
    raise ValueError('pad must be non-negative')
  cols = int(cols)
  assert cols >= 1
  N, H, W, C = imarray.shape
  rows = N // cols + int(N % cols != 0)
  batch_pad = rows * cols - N
  assert batch_pad >= 0
  post_pad = [batch_pad, pad, pad, 0]
  pad_arg = [[0, p] for p in post_pad]
  imarray = np.pad(imarray, pad_arg, 'constant', constant_values=padval)
  H += pad
  W += pad
  grid = (imarray
          .reshape(rows, cols, H, W, C)
          .transpose(0, 2, 1, 3, 4)
          .reshape(rows*H, cols*W, C))
  if pad:
    grid = grid[:-pad, :-pad]
  return grid

def interleave(*args):
  """Interleaves input arrays of the same shape along the batch axis."""
  if not args:
    raise ValueError('At least one argument is required.')
  a0 = args[0]
  if any(a.shape != a0.shape for a in args):
    raise ValueError('All inputs must have the same shape.')
  if not a0.shape:
    raise ValueError('Inputs must have at least one axis.')
  out = np.transpose(args, [1, 0] + list(range(2, len(a0.shape) + 1)))
  out = out.reshape(-1, *a0.shape[1:])
  return out

def imshow(a, format='png', jpeg_fallback=True):
  """Displays an image in the given format."""
  a = a.astype(np.uint8)
  data = io.BytesIO()
  PIL.Image.fromarray(a).save(data, format)
  im_data = data.getvalue()
  try:
    disp = IPython.display.display(IPython.display.Image(im_data))
  except IOError:
    if jpeg_fallback and format != 'jpeg':
      print ('Warning: image was too large to display in format "{}"; '
             'trying jpeg instead.').format(format)
      return imshow(a, format='jpeg')
    else:
      raise
  return disp

def image_to_uint8(x):
  """Converts [-1, 1] float array to [0, 255] uint8."""
  x = np.asarray(x)
  x = (256. / 2.) * (x + 1.)
  x = np.clip(x, 0, 255)
  x = x.astype(np.uint8)
  return x

Chargez un module BigBiGAN TF Hub et affichez ses fonctionnalités disponibles

# module = hub.Module(module_path, trainable=True, tags={'train'})  # training
module = hub.Module(module_path)  # inference

for signature in module.get_signature_names():
  print('Signature:', signature)
  print('Inputs:', pformat(module.get_input_info_dict(signature)))
  print('Outputs:', pformat(module.get_output_info_dict(signature)))
  print()
Signature: discriminate
Inputs: {'x': <hub.ParsedTensorInfo shape=(?, 128, 128, 3) dtype=float32 is_sparse=False>,
 'z': <hub.ParsedTensorInfo shape=(?, 120) dtype=float32 is_sparse=False>}
Outputs: {'score_x': <hub.ParsedTensorInfo shape=(?,) dtype=float32 is_sparse=False>,
 'score_xz': <hub.ParsedTensorInfo shape=(?,) dtype=float32 is_sparse=False>,
 'score_z': <hub.ParsedTensorInfo shape=(?,) dtype=float32 is_sparse=False>}

Signature: generate
Inputs: {'z': <hub.ParsedTensorInfo shape=(?, 120) dtype=float32 is_sparse=False>}
Outputs: {'default': <hub.ParsedTensorInfo shape=(?, 128, 128, 3) dtype=float32 is_sparse=False>,
 'upsampled': <hub.ParsedTensorInfo shape=(?, 256, 256, 3) dtype=float32 is_sparse=False>}

Signature: encode
Inputs: {'x': <hub.ParsedTensorInfo shape=(?, 256, 256, 3) dtype=float32 is_sparse=False>}
Outputs: {'avepool_feat': <hub.ParsedTensorInfo shape=(?, 2048) dtype=float32 is_sparse=False>,
 'bn_crelu_feat': <hub.ParsedTensorInfo shape=(?, 4096) dtype=float32 is_sparse=False>,
 'default': <hub.ParsedTensorInfo shape=(?, 120) dtype=float32 is_sparse=False>,
 'z_mean': <hub.ParsedTensorInfo shape=(?, 120) dtype=float32 is_sparse=False>,
 'z_sample': <hub.ParsedTensorInfo shape=(?, 120) dtype=float32 is_sparse=False>,
 'z_stdev': <hub.ParsedTensorInfo shape=(?, 120) dtype=float32 is_sparse=False>}

Signature: default
Inputs: {'x': <hub.ParsedTensorInfo shape=(?, 256, 256, 3) dtype=float32 is_sparse=False>}
Outputs: {'default': <hub.ParsedTensorInfo shape=(?, 120) dtype=float32 is_sparse=False>}

Définir une classe wrapper pour un accès pratique à diverses fonctions

class BigBiGAN(object):

  def __init__(self, module):
    """Initialize a BigBiGAN from the given TF Hub module."""
    self._module = module

  def generate(self, z, upsample=False):
    """Run a batch of latents z through the generator to generate images.

    Args:
      z: A batch of 120D Gaussian latents, shape [N, 120].

    Returns: a batch of generated RGB images, shape [N, 128, 128, 3], range
      [-1, 1].
    """
    outputs = self._module(z, signature='generate', as_dict=True)
    return outputs['upsampled' if upsample else 'default']

  def make_generator_ph(self):
    """Creates a tf.placeholder with the dtype & shape of generator inputs."""
    info = self._module.get_input_info_dict('generate')['z']
    return tf.placeholder(dtype=info.dtype, shape=info.get_shape())

  def gen_pairs_for_disc(self, z):
    """Compute generator input pairs (G(z), z) for discriminator, given z.

    Args:
      z: A batch of latents (120D standard Gaussians), shape [N, 120].

    Returns: a tuple (G(z), z) of discriminator inputs.
    """
    # Downsample 256x256 image x for 128x128 discriminator input.
    x = self.generate(z)
    return x, z

  def encode(self, x, return_all_features=False):
    """Run a batch of images x through the encoder.

    Args:
      x: A batch of data (256x256 RGB images), shape [N, 256, 256, 3], range
        [-1, 1].
      return_all_features: If True, return all features computed by the encoder.
        Otherwise (default) just return a sample z_hat.

    Returns: the sample z_hat of shape [N, 120] (or a dict of all features if
      return_all_features).
    """
    outputs = self._module(x, signature='encode', as_dict=True)
    return outputs if return_all_features else outputs['z_sample']

  def make_encoder_ph(self):
    """Creates a tf.placeholder with the dtype & shape of encoder inputs."""
    info = self._module.get_input_info_dict('encode')['x']
    return tf.placeholder(dtype=info.dtype, shape=info.get_shape())

  def enc_pairs_for_disc(self, x):
    """Compute encoder input pairs (x, E(x)) for discriminator, given x.

    Args:
      x: A batch of data (256x256 RGB images), shape [N, 256, 256, 3], range
        [-1, 1].

    Returns: a tuple (downsample(x), E(x)) of discriminator inputs.
    """
    # Downsample 256x256 image x for 128x128 discriminator input.
    x_down = tf.nn.avg_pool(x, ksize=2, strides=2, padding='SAME')
    z = self.encode(x)
    return x_down, z

  def discriminate(self, x, z):
    """Compute the discriminator scores for pairs of data (x, z).

    (x, z) must be batches with the same leading batch dimension, and joint
      scores are computed on corresponding pairs x[i] and z[i].

    Args:
      x: A batch of data (128x128 RGB images), shape [N, 128, 128, 3], range
        [-1, 1].
      z: A batch of latents (120D standard Gaussians), shape [N, 120].

    Returns:
      A dict of scores:
        score_xz: the joint scores for the (x, z) pairs.
        score_x: the unary scores for x only.
        score_z: the unary scores for z only.
    """
    inputs = dict(x=x, z=z)
    return self._module(inputs, signature='discriminate', as_dict=True)

  def reconstruct_x(self, x, use_sample=True, upsample=False):
    """Compute BigBiGAN reconstructions of images x via G(E(x)).

    Args:
      x: A batch of data (256x256 RGB images), shape [N, 256, 256, 3], range
        [-1, 1].
      use_sample: takes a sample z_hat ~ E(x). Otherwise, deterministically
        use the mean. (Though a sample z_hat may be far from the mean z,
        typically the resulting recons G(z_hat) and G(z) are very
        similar.
      upsample: if set, upsample the reconstruction to the input resolution
        (256x256). Otherwise return the raw lower resolution generator output
        (128x128).

    Returns: a batch of recons G(E(x)), shape [N, 256, 256, 3] if
      `upsample`, otherwise [N, 128, 128, 3].
    """
    if use_sample:
      z = self.encode(x)
    else:
      z = self.encode(x, return_all_features=True)['z_mean']
    recons = self.generate(z, upsample=upsample)
    return recons

  def losses(self, x, z):
    """Compute per-module BigBiGAN losses given data & latent sample batches.

    Args:
      x: A batch of data (256x256 RGB images), shape [N, 256, 256, 3], range
        [-1, 1].
      z: A batch of latents (120D standard Gaussians), shape [M, 120].

    For the original BigBiGAN losses, pass batches of size N=M=2048, with z's
    sampled from a 120D standard Gaussian (e.g., np.random.randn(2048, 120)),
    and x's sampled from the ImageNet (ILSVRC2012) training set with the
    "ResNet-style" preprocessing from:

        https://github.com/tensorflow/tpu/blob/master/models/official/resnet/resnet_preprocessing.py

    Returns:
      A dict of per-module losses:
        disc: loss for the discriminator.
        enc: loss for the encoder.
        gen: loss for the generator.
    """
    # Compute discriminator scores on (x, E(x)) pairs.
    # Downsample 256x256 image x for 128x128 discriminator input.
    scores_enc_x_dict = self.discriminate(*self.enc_pairs_for_disc(x))
    scores_enc_x = tf.concat([scores_enc_x_dict['score_xz'],
                              scores_enc_x_dict['score_x'],
                              scores_enc_x_dict['score_z']], axis=0)

    # Compute discriminator scores on (G(z), z) pairs.
    scores_gen_z_dict = self.discriminate(*self.gen_pairs_for_disc(z))
    scores_gen_z = tf.concat([scores_gen_z_dict['score_xz'],
                              scores_gen_z_dict['score_x'],
                              scores_gen_z_dict['score_z']], axis=0)

    disc_loss_enc_x = tf.reduce_mean(tf.nn.relu(1. - scores_enc_x))
    disc_loss_gen_z = tf.reduce_mean(tf.nn.relu(1. + scores_gen_z))
    disc_loss = disc_loss_enc_x + disc_loss_gen_z

    enc_loss = tf.reduce_mean(scores_enc_x)
    gen_loss = tf.reduce_mean(-scores_gen_z)

    return dict(disc=disc_loss, enc=enc_loss, gen=gen_loss)

Créez des tenseurs à utiliser plus tard pour calculer des échantillons, des reconstructions, des scores de discriminateur et des pertes

bigbigan = BigBiGAN(module)

# Make input placeholders for x (`enc_ph`) and z (`gen_ph`).
enc_ph = bigbigan.make_encoder_ph()
gen_ph = bigbigan.make_generator_ph()

# Compute samples G(z) from encoder input z (`gen_ph`).
gen_samples = bigbigan.generate(gen_ph)

# Compute reconstructions G(E(x)) of encoder input x (`enc_ph`).
recon_x = bigbigan.reconstruct_x(enc_ph, upsample=True)

# Compute encoder features used for representation learning evaluations given
# encoder input x (`enc_ph`).
enc_features = bigbigan.encode(enc_ph, return_all_features=True)

# Compute discriminator scores for encoder pairs (x, E(x)) given x (`enc_ph`)
# and generator pairs (G(z), z) given z (`gen_ph`).
disc_scores_enc = bigbigan.discriminate(*bigbigan.enc_pairs_for_disc(enc_ph))
disc_scores_gen = bigbigan.discriminate(*bigbigan.gen_pairs_for_disc(gen_ph))

# Compute losses.
losses = bigbigan.losses(enc_ph, gen_ph)
INFO:tensorflow:Saver not created because there are no variables in the graph to restore
INFO:tensorflow:Saver not created because there are no variables in the graph to restore
INFO:tensorflow:Saver not created because there are no variables in the graph to restore
INFO:tensorflow:Saver not created because there are no variables in the graph to restore
INFO:tensorflow:Saver not created because there are no variables in the graph to restore
INFO:tensorflow:Saver not created because there are no variables in the graph to restore
INFO:tensorflow:Saver not created because there are no variables in the graph to restore
INFO:tensorflow:Saver not created because there are no variables in the graph to restore
INFO:tensorflow:Saver not created because there are no variables in the graph to restore
INFO:tensorflow:Saver not created because there are no variables in the graph to restore
INFO:tensorflow:Saver not created because there are no variables in the graph to restore
INFO:tensorflow:Saver not created because there are no variables in the graph to restore
INFO:tensorflow:Saver not created because there are no variables in the graph to restore
INFO:tensorflow:Saver not created because there are no variables in the graph to restore
INFO:tensorflow:Saver not created because there are no variables in the graph to restore
INFO:tensorflow:Saver not created because there are no variables in the graph to restore
INFO:tensorflow:Saver not created because there are no variables in the graph to restore
INFO:tensorflow:Saver not created because there are no variables in the graph to restore
INFO:tensorflow:Saver not created because there are no variables in the graph to restore
INFO:tensorflow:Saver not created because there are no variables in the graph to restore
INFO:tensorflow:Saver not created because there are no variables in the graph to restore
INFO:tensorflow:Saver not created because there are no variables in the graph to restore
INFO:tensorflow:Saver not created because there are no variables in the graph to restore
INFO:tensorflow:Saver not created because there are no variables in the graph to restore

Créer une session TensorFlow et initialiser les variables

init = tf.global_variables_initializer()
sess = tf.Session()
sess.run(init)
2021-07-29 11:35:10.709616: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcuda.so.1
2021-07-29 11:35:11.349959: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2021-07-29 11:35:11.350872: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1733] Found device 0 with properties: 
pciBusID: 0000:00:05.0 name: Tesla V100-SXM2-16GB computeCapability: 7.0
coreClock: 1.53GHz coreCount: 80 deviceMemorySize: 15.78GiB deviceMemoryBandwidth: 836.37GiB/s
2021-07-29 11:35:11.350903: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudart.so.11.0
2021-07-29 11:35:11.355733: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcublas.so.11
2021-07-29 11:35:11.355835: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcublasLt.so.11
2021-07-29 11:35:11.357652: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcufft.so.10
2021-07-29 11:35:11.358067: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcurand.so.10
2021-07-29 11:35:11.360035: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcusolver.so.11
2021-07-29 11:35:11.361639: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcusparse.so.11
2021-07-29 11:35:11.361821: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudnn.so.8
2021-07-29 11:35:11.361937: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2021-07-29 11:35:11.362848: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2021-07-29 11:35:11.363716: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1871] Adding visible gpu devices: 0
2021-07-29 11:35:11.364273: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  AVX2 AVX512F FMA
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
2021-07-29 11:35:11.364791: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2021-07-29 11:35:11.365674: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1733] Found device 0 with properties: 
pciBusID: 0000:00:05.0 name: Tesla V100-SXM2-16GB computeCapability: 7.0
coreClock: 1.53GHz coreCount: 80 deviceMemorySize: 15.78GiB deviceMemoryBandwidth: 836.37GiB/s
2021-07-29 11:35:11.365794: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2021-07-29 11:35:11.366810: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2021-07-29 11:35:11.367699: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1871] Adding visible gpu devices: 0
2021-07-29 11:35:11.367762: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudart.so.11.0
2021-07-29 11:35:11.987148: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1258] Device interconnect StreamExecutor with strength 1 edge matrix:
2021-07-29 11:35:11.987192: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1264]      0 
2021-07-29 11:35:11.987201: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1277] 0:   N 
2021-07-29 11:35:11.987442: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2021-07-29 11:35:11.988506: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2021-07-29 11:35:11.989436: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2021-07-29 11:35:11.990378: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1418] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 14646 MB memory) -> physical GPU (device: 0, name: Tesla V100-SXM2-16GB, pci bus id: 0000:00:05.0, compute capability: 7.0)
2021-07-29 11:35:19.938556: I tensorflow/core/platform/profile_utils/cpu_utils.cc:114] CPU Frequency: 2000179999 Hz

Échantillons de générateur

Tout d' abord, nous visualisons des échantillons provenant du générateur BigBiGAN par échantillonnage pré - entraîné générateur entrées z d'une gaussienne standard (via np.random.randn ) et l' affichage des images qu'il produit. Jusqu'à présent, nous n'allons pas au-delà des capacités d'un GAN standard - nous utilisons simplement le générateur (et ignorons l'encodeur) pour le moment.

feed_dict = {gen_ph: np.random.randn(32, 120)}
_out_samples = sess.run(gen_samples, feed_dict=feed_dict)
print('samples shape:', _out_samples.shape)
imshow(imgrid(image_to_uint8(_out_samples), cols=4))
2021-07-29 11:35:26.648222: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcublas.so.11
2021-07-29 11:35:27.059336: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcublasLt.so.11
2021-07-29 11:35:27.068069: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudnn.so.8
2021-07-29 11:35:27.460253: I tensorflow/stream_executor/cuda/cuda_dnn.cc:359] Loaded cuDNN version 8100
samples shape: (32, 128, 128, 3)

png

Charge test_images de l'ensemble de données TF-Fleurs

BigBiGAN est formé sur ImageNet, mais comme il est trop volumineux pour fonctionner dans cette démo, nous utilisons le plus petit jeu de données TF-Flowers [1] comme entrée pour visualiser les reconstructions et calculer les fonctionnalités de l'encodeur.

Dans cette cellule on charge TF-Fleurs (téléchargement de l'ensemble de données si nécessaire) et stocker un lot fixe de 256x256 échantillons d'image RVB dans un tableau numpy test_images .

[1] https://www.tensorflow.org/datasets/catalog/tf_flowers

def get_flowers_data():
  """Returns a [32, 256, 256, 3] np.array of preprocessed TF-Flowers samples."""
  import tensorflow_datasets as tfds
  ds, info = tfds.load('tf_flowers', split='train', with_info=True)

  # Just get the images themselves as we don't need labels for this demo.
  ds = ds.map(lambda x: x['image'])

  # Filter out small images (with minor edge length <256).
  ds = ds.filter(lambda x: tf.reduce_min(tf.shape(x)[:2]) >= 256)

  # Take the center square crop of the image and resize to 256x256.
  def crop_and_resize(image):
    imsize = tf.shape(image)[:2]
    minor_edge = tf.reduce_min(imsize)
    start = (imsize - minor_edge) // 2
    stop = start + minor_edge
    cropped_image = image[start[0] : stop[0], start[1] : stop[1]]
    resized_image = tf.image.resize_bicubic([cropped_image], [256, 256])[0]
    return resized_image
  ds = ds.map(crop_and_resize)

  # Convert images from [0, 255] uint8 to [-1, 1] float32.
  ds = ds.map(lambda image: tf.cast(image, tf.float32) / (255. / 2.) - 1)

  # Take the first 32 samples.
  ds = ds.take(32)

  return np.array(list(tfds.as_numpy(ds)))

test_images = get_flowers_data()
2021-07-29 11:35:33.551693: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2021-07-29 11:35:33.552082: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1733] Found device 0 with properties: 
pciBusID: 0000:00:05.0 name: Tesla V100-SXM2-16GB computeCapability: 7.0
coreClock: 1.53GHz coreCount: 80 deviceMemorySize: 15.78GiB deviceMemoryBandwidth: 836.37GiB/s
2021-07-29 11:35:33.552227: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2021-07-29 11:35:33.552541: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2021-07-29 11:35:33.552809: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1871] Adding visible gpu devices: 0
2021-07-29 11:35:33.552846: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1258] Device interconnect StreamExecutor with strength 1 edge matrix:
2021-07-29 11:35:33.552853: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1264]      0 
2021-07-29 11:35:33.552860: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1277] 0:   N 
2021-07-29 11:35:33.552980: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2021-07-29 11:35:33.553282: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2021-07-29 11:35:33.553545: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1418] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 14646 MB memory) -> physical GPU (device: 0, name: Tesla V100-SXM2-16GB, pci bus id: 0000:00:05.0, compute capability: 7.0)
2021-07-29 11:35:33.724560: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1258] Device interconnect StreamExecutor with strength 1 edge matrix:
2021-07-29 11:35:33.724606: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1264]      
2021-07-29 11:35:44.049106: I tensorflow/compiler/mlir/mlir_graph_optimization_pass.cc:176] None of the MLIR Optimization Passes are enabled (registered 2)
2021-07-29 11:35:49.356913: W tensorflow/core/kernels/data/cache_dataset_ops.cc:768] The calling iterator did not fully read the dataset being cached. In order to avoid unexpected truncation of the dataset, the partially cached contents of the dataset  will be discarded. This can happen if you have an input pipeline similar to `dataset.cache().take(k).repeat()`. You should use `dataset.take(k).cache().repeat()` instead.

Reconstitutions

Maintenant , nous visualisons reconstructions BigBiGAN en passant par des images réelles du codeur et à travers le générateur, calcul G(E(x)) images données x . Ci - dessous, les images d' entrée x sont indiquées dans la colonne de gauche, et les reconstructions correspondantes sont indiquées sur la droite.

Notez que les reconstructions ne sont pas des correspondances parfaites au pixel près avec les images d'entrée ; ils ont plutôt tendance à capturer le contenu sémantique de niveau supérieur de l'entrée tout en "oubliant" la plupart des détails de bas niveau. Cela suggère que l'encodeur BigBiGAN peut apprendre à capturer les types d'informations sémantiques de haut niveau sur les images que nous aimerions voir dans une approche d'apprentissage de représentation.

Notez également que les reconstructions brutes des images d'entrée 256x256 sont à la résolution inférieure produite par notre générateur - 128x128. Nous les suréchantillonnons à des fins de visualisation.

test_images_batch = test_images[:16]
_out_recons = sess.run(recon_x, feed_dict={enc_ph: test_images_batch})
print('reconstructions shape:', _out_recons.shape)

inputs_and_recons = interleave(test_images_batch, _out_recons)
print('inputs_and_recons shape:', inputs_and_recons.shape)
imshow(imgrid(image_to_uint8(inputs_and_recons), cols=2))
reconstructions shape: (16, 256, 256, 3)
inputs_and_recons shape: (32, 256, 256, 3)

png

Fonctionnalités de l'encodeur

Nous montrons maintenant comment calculer des caractéristiques à partir de l'encodeur utilisé pour les évaluations d'apprentissage de représentation standard.

Ces caractéristiques pourraient être utilisées dans un classificateur linéaire ou basé sur les voisins les plus proches. Nous incluons la fonction standard prise après la mise en commun moyenne mondiale (clé avepool_feat ) ainsi que la plus grande caractéristique « BN + CReLU » (touche bn_crelu_feat ) utilisé pour obtenir les meilleurs résultats.

_out_features = sess.run(enc_features, feed_dict={enc_ph: test_images_batch})
print('AvePool features shape:', _out_features['avepool_feat'].shape)
print('BN+CReLU features shape:', _out_features['bn_crelu_feat'].shape)
AvePool features shape: (16, 2048)
BN+CReLU features shape: (16, 4096)

Scores et pertes du discriminateur

Enfin, nous calculerons les scores et les pertes du discriminateur sur des lots de paires de codeurs et de générateurs. Ces pertes pourraient être transmises à un optimiseur pour entraîner BigBiGAN.

Nous utilisons notre lot d'images ci - dessus comme les entrées du codeur x , calculer le score de codeur D(x, E(x)) . Pour les entrées du générateur nous échantillon z à partir d' une norme gaussienne via 120D np.random.randn , le calcul du score de générateur en tant que D(G(z), z) .

Le discriminateur prédit un score commun score_xz pour les (x, z) paires ainsi que les scores unaires score_x et score_z pour x et z seuls, respectivement. Il est entraîné à donner des scores élevés (positifs) aux paires d'encodeurs et des scores faibles (négatifs) aux paires de générateurs. Cela est la plupart du temps ci - dessous, bien que la unaire score_z est négative dans les deux cas, ce qui indique que le codeur sorties E(x) ressemblent à des échantillons réels provenant d' une gaussienne.

feed_dict = {enc_ph: test_images, gen_ph: np.random.randn(32, 120)}
_out_scores_enc, _out_scores_gen, _out_losses = sess.run(
    [disc_scores_enc, disc_scores_gen, losses], feed_dict=feed_dict)
print('Encoder scores:', {k: v.mean() for k, v in _out_scores_enc.items()})
print('Generator scores:', {k: v.mean() for k, v in _out_scores_gen.items()})
print('Losses:', _out_losses)
Encoder scores: {'score_z': -0.50417066, 'score_xz': 0.6934861, 'score_x': 1.4621685}
Generator scores: {'score_z': -0.4306627, 'score_xz': -0.76503456, 'score_x': -0.5694851}
Losses: {'disc': 1.2889439, 'enc': 0.54947025, 'gen': 0.5883941}