We have released pretrained BigBiGAN models used for unsupervised image generation and representation learning, as described in our July 2019 tech report, "Large Scale Adversarial Representation Learning" [1].
The pretrained models are available for use via TF Hub. The release includes two BigBiGAN models with different encoder architectures:
- Small encoder (ResNet-50): https://tfhub.dev/deepmind/bigbigan-resnet50/1
- Large encoder (RevNet-50 x4): https://tfhub.dev/deepmind/bigbigan-revnet50x4/1
See the TF Hub pages linked above for documentation and example usage of each module.
A Google Colab-based demo with example usage of the model functionality and sample visualization is available here.
The snippet below demonstrates the use of the released TF Hub modules for image generation/reconstruction and encoder feature computation. (The Colab demo includes more extensive documentation and visualizations.)
# Load BigBiGAN module.
module = hub.Module('https://tfhub.dev/deepmind/bigbigan-resnet50/1') # small encoder
# module = hub.Module('https://tfhub.dev/deepmind/bigbigan-revnet50x4/1') # large encoder
# Sample a batch of 8 random latent vectors (z) from the Gaussian prior. Then
# call the generator on the latent samples to generate a batch of images with
# shape [8, 128, 128, 3] and range [-1, 1].
z = tf.random.normal([8, 120]) # latent samples
gen_samples = module(z, signature='generate')
# Given a batch of 256x256 RGB images in range [-1, 1], call the encoder to
# compute predicted latents z and other features (e.g. for use in downstream
# recognition tasks).
images = tf.placeholder(tf.float32, shape=[None, 256, 256, 3])
features = module(images, signature='encode', as_dict=True)
# Get the predicted latent sample `z_sample` from the dict of features.
# Other available features include `avepool_feat` and `bn_crelu_feat`, used in
# the representation learning results.
z_sample = features['z_sample'] # shape [?, 120]
# Compute reconstructions of the input `images` by passing the encoder's output
# `z_sample` back through the generator. Note that raw generator outputs are
# half the resolution of encoder inputs (128x128). To get upsampled generator
# outputs matching the encoder input resolution (256x256), instead use:
# recons = module(z_sample, signature='generate', as_dict=True)['upsampled']
recons = module(z_sample, signature='generate') # shape [?, 128, 128, 3][1] Jeff Donahue and Karen Simonyan. Large Scale Adversarial Representation Learning. arxiv:1907.02544, 2019.