Rethinking the Truly Unsupervised Image-to-Image Translation
Authors:
- 1 October 2021
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 14134-14143
Abstract
Every recent image-to-image translation model inherently requires either image-level (i.e. input-output pairs) or set-level (i.e. domain labels) supervision. However, even set-level supervision can be a severe bottleneck for data collection in practice. In this paper, we tackle image-to-image translation in a fully unsupervised setting, i.e., neither paired images nor domain labels. To this end, we propose a truly unsupervised image-to-image translation model (TUNIT) that simultaneously learns to separate image domains and translates input images into the estimated domains. Experimental results show that our model achieves comparable or even better performance than the set-level supervised model trained with full labels, generalizes well on various datasets, and is robust against the choice of hyperparameters (e.g. the preset number of pseudo domains). Furthermore, TUNIT can be easily extended to semi-supervised learning with a few labeled data.Keywords
All Related Versions
This publication has 15 references indexed in Scilit:
- DRIT++: Diverse Image-to-Image Translation via Disentangled RepresentationsInternational Journal of Computer Vision, 2020
- Few-Shot Unsupervised Image-to-Image Translation2019
- Invariant Information Clustering for Unsupervised Image Classification and Segmentation2019
- Semantic Image Synthesis With Spatially-Adaptive Normalization2019
- A Style-Based Generator Architecture for Generative Adversarial Networks2019
- High-Resolution Image Synthesis and Semantic Manipulation with Conditional GANs2018
- DeblurGAN: Blind Motion Deblurring Using Conditional Adversarial Networks2018
- StarGAN: Unified Generative Adversarial Networks for Multi-domain Image-to-Image Translation2018
- Deep Residual Learning for Image Recognition2016
- Dimensionality Reduction by Learning an Invariant Mapping2006