Learning What and Where to Draw

TL;DR

The paper introduces GAWWN, a GAN-based model that synthesizes images with control over content and spatial location, enabling detailed and flexible image generation from text and location instructions.

Contribution

We propose GAWWN, a novel GAN architecture that allows explicit control over what to draw and where, including part-level conditioning and flexible subset control.

Findings

High-quality 128x128 bird images conditioned on text and location

Control over object bounding boxes and parts in generated images

Preliminary results on human action image synthesis with location control

Abstract

Generative Adversarial Networks (GANs) have recently demonstrated the capability to synthesize compelling real-world images, such as room interiors, album covers, manga, faces, birds, and flowers. While existing models can synthesize images based on global constraints such as a class label or caption, they do not provide control over pose or object location. We propose a new model, the Generative Adversarial What-Where Network (GAWWN), that synthesizes images given instructions describing what content to draw in which location. We show high-quality 128 x 128 image synthesis on the Caltech-UCSD Birds dataset, conditioned on both informal text descriptions and also object location. Our system exposes control over both the bounding box around the bird and its constituent parts. By modeling the conditional distributions over part locations, our system also enables conditioning on arbitrary subsets of parts (e.g. only the beak and tail), yielding an efficient interface for picking part locations. We also show preliminary results on the more challenging domain of text- and location-controllable synthesis of images of human actions on the MPII Human Pose dataset.