TSIT: A Simple and Versatile Framework for Image-to-Image Translation
arXiv:2007.12072
Abstract
We introduce a simple and versatile framework for image-to-image translation. We unearth the importance of normalization layers, and provide a carefully designed two-stream generative model with newly proposed feature transformations in a coarse-to-fine fashion. This allows multi-scale semantic structure information and style representation to be effectively captured and fused by the network, permitting our method to scale to various tasks in both unsupervised and supervised settings. No additional constraints (e.g., cycle consistency) are needed, contributing to a very clean and simple method. Multi-modal image synthesis with arbitrary style control is made possible. A systematic study compares the proposed method with several state-of-the-art task-specific baselines, verifying its effectiveness in both perceptual quality and quantitative evaluations.
ECCV 2020 (Spotlight). Table 2 is updated. GitHub: https://github.com/EndlessSora/TSIT
References in corpus (9)
- Very Deep Convolutional Networks for Large-Scale Image Recognition
- Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
- Conditional Generative Adversarial Nets
- GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium
- Semi-Supervised Learning with Deep Generative Models
- A Learned Representation For Artistic Style
- Unsupervised Cross-Domain Image Generation
- Learning to Predict Layout-to-image Conditional Convolutions for Semantic Image Synthesis
- ETNet: Error Transition Network for Arbitrary Style Transfer