CAFLOW: Conditional Autoregressive Flows

TL;DR

CAFLOW is a novel image-to-image translation model that combines auto-regressive modeling with conditional normalizing flows, enabling diverse and high-quality image synthesis across multiple tasks.

Contribution

It introduces a multi-scale normalizing flow framework that models conditional distributions of latent encodings, enhancing flexibility and performance over previous conditional flow models.

Findings

Outperforms previous conditional flow models in image translation tasks.

Effectively models complex conditional distributions at multiple resolutions.

Demonstrates versatility across various image-to-image translation applications.

Abstract

We introduce CAFLOW, a new diverse image-to-image translation model that simultaneously leverages the power of auto-regressive modeling and the modeling efficiency of conditional normalizing flows. We transform the conditioning image into a sequence of latent encodings using a multi-scale normalizing flow and repeat the process for the conditioned image. We model the conditional distribution of the latent encodings by modeling the auto-regressive distributions with an efficient multi-scale normalizing flow, where each conditioning factor affects image synthesis at its respective resolution scale. Our proposed framework performs well on a range of image-to-image translation tasks. It outperforms former designs of conditional flows because of its expressive auto-regressive structure.