HiLLoC: Lossless Image Compression with Hierarchical Latent Variable Models

By James Townsend, Thomas Bird, Julius Kunze, David Barber. 2019

61 citations recorded by the source corpus.

We make the following striking observation: fully convolutional VAE models trained on 32x32 ImageNet can generalize well, not just to 64x64 but also to far larger photographs, with no changes to the model. We use this property, applying fully convolutional models to lossless compression, demonstrating a method to scale the VAE-based 'Bits-Back with ANS' algorithm for lossless compression to large color photographs, and achieving state of the art for compression of full size ImageNet images. We release Craystack, an open source library for convenient prototyping of lossless compression using probabilistic models, along with full implementations of all of our compression results.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Generative Models, Information Theory

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Generative Models

In Information Theory

Citations in the corpus

Fast Lossless Neural Compression with Integer-Only Discrete Flows (2022, background): “With a model distribution close enough to the data distribution, many efﬁcient lossless neural compressors have been explored (Hoogeboom et al., 2019; van den Berg et al., 2020; Ho et al., 2019b; Townsend et al., 2019a; Kingma et al., 2021).”
iFlow: Numerically Invertible Flows for Efficient Lossless Compression via a Uniform Coder (2021, methodology): “Alternatively, variational autoencoders (VAEs) [24, 26, 37] maximise a lower bound on the marginal data likelihood (otherwise known as the ELBO).”
Reducing the Computational Cost of Deep Generative Models with Binary Neural Networks (2020, background): “…useful result for any real-world application of these generative models. such as learned lossless compression (Townsend et al., 2019; Kingma et al., 2019; Townsend et al., 2020; Ho et al., 2019b; Hoogeboom et al., 2019), which could be made practical with these reduced memory and compute benefits.”
Improving Inference for Neural Image Compression (2020, methodology): “Recent work has improved its efficiency in hierarchical latent variable models [Townsend et al., 2019b, Kingma et al., 2019], and extended it to flow models [Ho et al., 2019].”
IDF++: Analyzing and Improving Integer Discrete Flows for Lossless Compression (2020, background): “More work in directions like optimizing for (de)compression speed [7] or generalizing learnable compressors to other datasets [50] is needed to make deep density estimators more practical for source compression.”
iVPF: Numerical Invertible Volume Preserving Flow for Efficient Lossless Compression (2021, methodology): “Like general lossless compression algorithms [15, 3, 36, 13], we use CIFAR10, ImageNet32 and ImageNet64 for training and evaluation.”
QARV: Quantization-Aware ResNet VAE for Lossy Image Compression (2023, methodology): “For example, recent studies have shown that many powerful generative models, such as auto-regressive models [1], Variational Autoencoders [2], Normalizing Flows [3], and diffusion probabilistic models [4], can also serve as highly effective lossless compressors.”