LR-GAN: Layered Recursive Generative Adversarial Networks for Image Generation

2017/03/05 by Jianwei Yang, Anitha Kannan, Yang, Jianwei +5 · 4 citations
Computer Science · #Computer Graphics and Visualization Techniques #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Generative Adversarial Networks and Image Synthesis #Machine Learning (cs.LG)

paper · pdf · doi:10.48550/arxiv.1703.01560

openalex publication_date 2017/03/05 · openalex created_date 2017/03/16 · openalex updated_date 2026/07/28

Abstract

We present LR-GAN: an adversarial image generation model which takes scene structure and context into account. Unlike previous generative adversarial networks (GANs), the proposed GAN learns to generate image background and foregrounds separately and recursively, and stitch the foregrounds on the background in a contextually relevant manner to produce a complete natural image. For each foreground, the model learns to generate its appearance, shape and pose. The whole model is unsupervised, and is trained in an end-to-end manner with gradient descent methods. The experiments demonstrate that LR-GAN can generate more natural images with objects that are more human recognizable than DCGAN.

Cited by

• Multi-Agent Amodal Completion: Direct Synthesis with Fine-Grained Semantic Guidance
• Spatial Fusion GAN for Image Synthesis
• Emergence of Object Segmentation in Perturbed Generative Models
• GaussiGAN: Controllable Image Synthesis with 3D Gaussians from Unposed Silhouettes

Related

• Interpreting Generative Adversarial Networks for Interactive Image Generation
• Systematic Analysis of Image Generation using GANs
• Generating Images Part by Part with Composite Generative Adversarial Networks
• Generative Adversarial Networks: An Overview
• Generative Image Modeling using Style and Structure Adversarial Networks