Neural Photo Editing with Introspective Adversarial Networks

Abstract

The increasingly photorealistic sample quality of generative image models suggests their feasibility in applications beyond image generation. We present the Neural Photo Editor, an interface that leverages the power of generative neural networks to make large, semantically coherent changes to existing images. To tackle the challenge of achieving accurate reconstructions without loss of feature quality, we introduce the Introspective Adversarial Network, a novel hybridization of the VAE and GAN. Our model efficiently captures long-range dependencies through use of a computational block based on weight-shared dilated convolutions, and improves generalization performance with Orthogonal Regularization, a novel weight regularization method. We validate our contributions on CelebA, SVHN, and CIFAR-100, and produce samples and reconstructions with high visual fidelity.

10 pages, 7 figures, 3 tables

References in corpus (1)

• An All-In-One Convolutional Neural Network for Face Analysis

Cited by in corpus (19)

• Unsupervised Cross-Domain Image Generation
• A Review on Generative Adversarial Networks: Algorithms, Theory, and Applications
• Optimal transport maps for distribution preserving operations on latent spaces of Generative Models
• RankGAN: A Maximum Margin Ranking GAN for Generating Faces
• Jigsaw-VAE: Towards Balancing Features in Variational Autoencoders
• Decomposed Adversarial Learned Inference
• Depth Structure Preserving Scene Image Generation
• Approximated Orthonormal Normalisation in Training Neural Networks
• Exploiting GAN Internal Capacity for High-Quality Reconstruction of Natural Images
• Intuitive, Interactive Beard and Hair Synthesis with Generative Models
• Semantic Image Manipulation Using Scene Graphs
• A Generative Approach Towards Improved Robotic Detection of Marine Litter
• Inverse mapping of face GANs
• Kernel Mean Matching for Content Addressability of GANs
• D2C: Diffusion-Denoising Models for Few-shot Conditional Generation
• MakeupBag: Disentangling Makeup Extraction and Application
• Adversarial Code Learning for Image Generation
• Diversity Regularized Adversarial Learning
• One-element Batch Training by Moving Window