Towards Instance-level Image-to-Image Translation - Shen - CVPR 2019

 

Info

  • Title: Towards Instance-level Image-to-Image Translation
  • Task: Image Translation
  • Author: Zhiqiang Shen, Mingyang Huang, Jianping Shi, Xiangyang Xue, Thomas Huang
  • Date: May 2019
  • Arxiv: 1905.01744
  • Published: CVPR 2019

Highlights & Drawbacks

  • The instance-level objective loss can help learn a more accurate reconstruction and incorporate diverse attributes of objects
  • A more reasonable mapping: the styles used for target domain of local/global areas are from corresponding spatial regions in source domain.
  • A large-scale, multimodal, highly varied Image-to-Image translation dataset, containing ∼155k streetscape images across four domains.

Motivation & Design

Disentangle background and object style in translation process: Towards Instance-level Image-to-Image Translation

The framework overview: Towards Instance-level Image-to-Image Translation

Loss Design Towards Instance-level Image-to-Image Translation

Towards Instance-level Image-to-Image Translation

The instance-level translation dataset and comparisons with previous ones: Towards Instance-level Image-to-Image Translation

Performance & Ablation Study

Towards Instance-level Image-to-Image Translation

The authors compared results with baselines like UNIT, CycleGAN, MUNIT and DRIT, using LPIPS distance to measure the diversity of generated images.

A visualization for generated style distribution is also provided.

Code

Project Site