Natural Image Manipulation for Autoregressive Models Using Fisher Scores

Wilson Yan, Jonathan Ho, Pieter Abbeel

Keywords: fisher score, generative models, image interpolation

TL;DR: We develop a novel method to perform image interpolation and semantic manipulation using autoregressive models through fisher scores

Abstract: Deep autoregressive models are one of the most powerful models that exist today which achieve state-of-the-art bits per dim. However, they lie at a strict disadvantage when it comes to controlled sample generation compared to latent variable models. Latent variable models such as VAEs and normalizing flows allow meaningful semantic manipulations in latent space, which autoregressive models do not have. In this paper, we propose using Fisher scores as a method to extract embeddings from an autoregressive model to use for interpolation and show that our method provides more meaningful sample manipulation compared to alternate embeddings such as network activations.

Community Implementations: [ 1 code implementation](https://www.catalyzex.com/paper/arxiv:1912.05015/code)