StyleGAN2 Distillation for Feed-forward Image Manipulation

Abstract

StyleGAN2 is a state-of-the-art network in generating realistic images. Besides, it was explicitly trained to have disentangled directions in latent space, which allows efficient image manipulation by varying latent factors. Editing existing images requires embedding a given image into the latent space of StyleGAN2. Latent code optimization via backpropagation is commonly used for qualitative embedding of real world images, although it is prohibitively slow for many applications. We propose a way to distill a particular image manipulation of StyleGAN2 into image-to-image network trained in paired way. The resulting pipeline is an alternative to existing GANs, trained on unpaired data. We provide results of human faces' transformation: gender swap, aging/rejuvenation, style transfer and image morphing. We show that the quality of generation using our method is comparable to StyleGAN2 backpropagation and current state-of-the-art methods in these particular tasks.

arXiv commentCamera ready ECCV 2020

References in corpus (15)

• Distilling the Knowledge in a Neural Network
• Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
• GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium
• Large Scale GAN Training for High Fidelity Natural Image Synthesis
• Progressive Growing of GANs for Improved Quality, Stability, and Variation
• Toward Multimodal Image-to-Image Translation
• Self-training with Noisy Student improves ImageNet classification
• Classification Accuracy Score for Conditional Generative Models
• Video-to-Video Synthesis
• StarGAN v2: Diverse Image Synthesis for Multiple Domains
• Compressing GANs using Knowledge Distillation
• Self-ensembling for visual domain adaptation
• On the "steerability" of generative adversarial networks
• Style Generator Inversion for Image Enhancement and Animation
• Learning To Simulate

Cited by in corpus (9)

• PVA: Pixel-aligned Volumetric Avatars
• Only a Matter of Style: Age Transformation Using a Style-Based Regression Model
• MorphGAN: One-Shot Face Synthesis GAN for Detecting Recognition Bias
• Enjoy Your Editing: Controllable GANs for Image Editing via Latent Space Navigation
• Powers of layers for image-to-image translation
• Distribution Aligned Multimodal and Multi-Domain Image Stylization
• Content-Aware GAN Compression
• Ensembling with Deep Generative Views
• Neural Crossbreed: Neural Based Image Metamorphosis