Enjoy Your Editing: Controllable GANs for Image Editing via Latent Space Navigation

TL;DR

This paper introduces a controllable GAN-based method for image editing that improves attribute disentanglement, preserves image identity, and enhances photo-realism through simultaneous attribute transformations and specialized training losses.

Contribution

It presents a novel approach that learns multiple attribute transformations together, incorporates attribute regression, and uses content and adversarial losses for better control and realism.

Findings

Achieves state-of-the-art performance in targeted image manipulation.

Effectively preserves image identity and realism during editing.

Provides quantitative evaluation strategies for controllable editing.

Abstract

Controllable semantic image editing enables a user to change entire image attributes with a few clicks, e.g., gradually making a summer scene look like it was taken in winter. Classic approaches for this task use a Generative Adversarial Net (GAN) to learn a latent space and suitable latent-space transformations. However, current approaches often suffer from attribute edits that are entangled, global image identity changes, and diminished photo-realism. To address these concerns, we learn multiple attribute transformations simultaneously, integrate attribute regression into the training of transformation functions, and apply a content loss and an adversarial loss that encourages the maintenance of image identity and photo-realism. We propose quantitative evaluation strategies for measuring controllable editing performance, unlike prior work, which primarily focuses on qualitative evaluation. Our model permits better control for both single- and multiple-attribute editing while preserving image identity and realism during transformation. We provide empirical results for both natural and synthetic images, highlighting that our model achieves state-of-the-art performance for targeted image manipulation.