Model-Based Occlusion Disentanglement for Image-to-Image Translation

Abstract

Image-to-image translation is affected by entanglement phenomena, which may occur in case of target data encompassing occlusions such as raindrops, dirt, etc. Our unsupervised model-based learning disentangles scene and occlusions, while benefiting from an adversarial pipeline to regress physical parameters of the occlusion model. The experiments demonstrate our method is able to handle varying types of occlusions and generate highly realistic translations, qualitatively and quantitatively outperforming the state-of-the-art on multiple datasets.

Cite

Text

Pizzati et al. "Model-Based Occlusion Disentanglement for Image-to-Image Translation." Proceedings of the European Conference on Computer Vision (ECCV), 2020. doi:10.1007/978-3-030-58565-5_27

Markdown

[Pizzati et al. "Model-Based Occlusion Disentanglement for Image-to-Image Translation." Proceedings of the European Conference on Computer Vision (ECCV), 2020.](https://mlanthology.org/eccv/2020/pizzati2020eccv-modelbased/) doi:10.1007/978-3-030-58565-5_27

BibTeX

@inproceedings{pizzati2020eccv-modelbased,
  title     = {{Model-Based Occlusion Disentanglement for Image-to-Image Translation}},
  author    = {Pizzati, Fabio and Cerri, Pietro and de Charette, Raoul},
  booktitle = {Proceedings of the European Conference on Computer Vision (ECCV)},
  year      = {2020},
  doi       = {10.1007/978-3-030-58565-5_27},
  url       = {https://mlanthology.org/eccv/2020/pizzati2020eccv-modelbased/}
}