Thanks for tuning in to Google I/O. View all sessions on demandWatch on demand

Generating Images with Little Data Using S3GAN

View on TensorFlow.org Run in Google Colab View on GitHub Download notebook See TF Hub models

This notebook is a demo of Generative Adversarial Networks trained on ImageNet with as little as 2.5% labeled data using self- and semi-supervised learning techniques. Both generator and discriminator models are available on TF Hub.

For more information about the models and the training procedure see our blogpost and the paper [1]. The code for training these models is available on GitHub.

To get started, connect to a runtime and follow these steps:

  1. (Optional) Select a model in the second code cell below.
  2. Click Runtime > Run all to run each cell in order.
    • Afterwards, the interactive visualizations should update automatically when you modify the settings using the sliders and dropdown menus.

[1] Mario Lucic*, Michael Tschannen*, Marvin Ritter*, Xiaohua Zhai, Olivier Bachem, Sylvain Gelly, High-Fidelity Image Generation With Fewer Labels, ICML 2019.

Setup

# @title Imports and utility functions
import os

import IPython
from IPython.display import display
import numpy as np
import PIL.Image
import pandas as pd
import six

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

import tensorflow_hub as hub

def imgrid(imarray, cols=8, pad=1):
  pad = int(pad)
  assert pad >= 0
  cols = int(cols)
  assert cols >= 1
  N, H, W, C = imarray.shape
  rows = int(np.ceil(N / float(cols)))
  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')
  H += pad
  W += pad
  grid = (imarray
          .reshape(rows, cols, H, W, C)
          .transpose(0, 2, 1, 3, 4)
          .reshape(rows*H, cols*W, C))
  return grid[:-pad, :-pad]


def imshow(a, format='png', jpeg_fallback=True):
  a = np.asarray(a, dtype=np.uint8)
  if six.PY3:
    str_file = six.BytesIO()
  else:
    str_file = six.StringIO()
  PIL.Image.fromarray(a).save(str_file, format)
  png_data = str_file.getvalue()
  try:
    disp = display(IPython.display.Image(png_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


class Generator(object):

  def __init__(self, module_spec):
    self._module_spec = module_spec
    self._sess = None
    self._graph = tf.Graph()
    self._load_model()

  @property
  def z_dim(self):
    return self._z.shape[-1].value

  @property
  def conditional(self):
    return self._labels is not None

  def _load_model(self):
    with self._graph.as_default():
      self._generator = hub.Module(self._module_spec, name="gen_module",
                                   tags={"gen", "bsNone"})
      input_info = self._generator.get_input_info_dict()
      inputs = {k: tf.placeholder(v.dtype, v.get_shape().as_list(), k)
                for k, v in self._generator.get_input_info_dict().items()}
      self._samples = self._generator(inputs=inputs, as_dict=True)["generated"]
      print("Inputs:", inputs)
      print("Outputs:", self._samples)
      self._z = inputs["z"]
      self._labels = inputs.get("labels", None)

  def _init_session(self):
    if self._sess is None:
      self._sess = tf.Session(graph=self._graph)
      self._sess.run(tf.global_variables_initializer())

  def get_noise(self, num_samples, seed=None):
    if np.isscalar(seed):
      np.random.seed(seed)
      return np.random.normal(size=[num_samples, self.z_dim])
    z = np.empty(shape=(len(seed), self.z_dim), dtype=np.float32)
    for i, s in enumerate(seed):
      np.random.seed(s)
      z[i] = np.random.normal(size=[self.z_dim])
    return z

  def get_samples(self, z, labels=None):
    with self._graph.as_default():
      self._init_session()
      feed_dict = {self._z: z}
      if self.conditional:
        assert labels is not None
        assert labels.shape[0] == z.shape[0]
        feed_dict[self._labels] = labels
      samples = self._sess.run(self._samples, feed_dict=feed_dict)
      return np.uint8(np.clip(256 * samples, 0, 255))


class Discriminator(object):

  def __init__(self, module_spec):
    self._module_spec = module_spec
    self._sess = None
    self._graph = tf.Graph()
    self._load_model()

  @property
  def conditional(self):
    return "labels" in self._inputs

  @property
  def image_shape(self):
    return self._inputs["images"].shape.as_list()[1:]

  def _load_model(self):
    with self._graph.as_default():
      self._discriminator = hub.Module(self._module_spec, name="disc_module",
                                       tags={"disc", "bsNone"})
      input_info = self._discriminator.get_input_info_dict()
      self._inputs = {k: tf.placeholder(v.dtype, v.get_shape().as_list(), k)
                      for k, v in input_info.items()}
      self._outputs = self._discriminator(inputs=self._inputs, as_dict=True)
      print("Inputs:", self._inputs)
      print("Outputs:", self._outputs)

  def _init_session(self):
    if self._sess is None:
      self._sess = tf.Session(graph=self._graph)
      self._sess.run(tf.global_variables_initializer())

  def predict(self, images, labels=None):
    with self._graph.as_default():
      self._init_session()
      feed_dict = {self._inputs["images"]: images}
      if "labels" in self._inputs:
        assert labels is not None
        assert labels.shape[0] == images.shape[0]
        feed_dict[self._inputs["labels"]] = labels
      return self._sess.run(self._outputs, feed_dict=feed_dict)

WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow/python/compat/v2_compat.py:107: 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

Select a model

# @title Select a model { run: "auto" }

model_name = "S3GAN 128x128 20% labels (FID 6.9, IS 98.1)"  # @param ["S3GAN 256x256 10% labels (FID 8.8, IS 130.7)", "S3GAN 128x128 2.5% labels (FID 12.6, IS 48.7)", "S3GAN 128x128 5% labels (FID 8.4, IS 74.0)", "S3GAN 128x128 10% labels (FID 7.6, IS 90.3)", "S3GAN 128x128 20% labels (FID 6.9, IS 98.1)"]
models = {
    "S3GAN 256x256 10% labels": "https://tfhub.dev/google/compare_gan/s3gan_10_256x256/1",
    "S3GAN 128x128 2.5% labels": "https://tfhub.dev/google/compare_gan/s3gan_2_5_128x128/1",
    "S3GAN 128x128 5% labels": "https://tfhub.dev/google/compare_gan/s3gan_5_128x128/1",
    "S3GAN 128x128 10% labels": "https://tfhub.dev/google/compare_gan/s3gan_10_128x128/1",
    "S3GAN 128x128 20% labels": "https://tfhub.dev/google/compare_gan/s3gan_20_128x128/1",
}

module_spec = models[model_name.split(" (")[0]]
print("Module spec:", module_spec)

tf.reset_default_graph()
print("Loading model...")
sampler = Generator(module_spec)
print("Model loaded.")

Module spec: https://tfhub.dev/google/compare_gan/s3gan_20_128x128/1
Loading model...
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
Inputs: {'z': <tf.Tensor 'z:0' shape=(?, 120) dtype=float32>, 'labels': <tf.Tensor 'labels:0' shape=(?,) dtype=int32>}
Outputs: Tensor("gen_module_apply_default/generator_1/truediv:0", shape=(?, 128, 128, 3), dtype=float32)
Model loaded.

Sample

png

png

Discriminator

disc = Discriminator(module_spec)

batch_size = 4
num_classes = 1000
images = np.random.random(size=[batch_size] + disc.image_shape)
labels = np.random.randint(0, num_classes, size=(batch_size))

disc.predict(images, labels=labels)

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
Inputs: {'images': <tf.Tensor 'images:0' shape=(?, 128, 128, 3) dtype=float32>, 'labels': <tf.Tensor 'labels:0' shape=(?,) dtype=int32>}
Outputs: {'prediction': <tf.Tensor 'disc_module_apply_default/discriminator/Sigmoid:0' shape=(?, 1) dtype=float32>}
{'prediction': array([[0.82321626],
        [0.890308  ],
        [0.86215377],
        [0.88563377]], dtype=float32)}