DRAW: A Recurrent Neural Network For Image Generation

TL;DR

DRAW is a neural network architecture that uses a recurrent, attention-based approach to generate high-quality images through iterative refinement, significantly advancing image generation capabilities.

Contribution

It introduces a novel spatial attention mechanism combined with a sequential variational auto-encoding framework for improved image synthesis.

Findings

Outperforms previous models on MNIST dataset

Generates highly realistic images on Street View House Numbers dataset

Achieves indistinguishability from real data in generated images

Abstract

This paper introduces the Deep Recurrent Attentive Writer (DRAW) neural network architecture for image generation. DRAW networks combine a novel spatial attention mechanism that mimics the foveation of the human eye, with a sequential variational auto-encoding framework that allows for the iterative construction of complex images. The system substantially improves on the state of the art for generative models on MNIST, and, when trained on the Street View House Numbers dataset, it generates images that cannot be distinguished from real data with the naked eye.