CoMoGAN: continuous model-guided image-to-image translation
Preprint
Abstract
CoMoGAN is a continuous GAN relying on the unsupervised reorganization of the target data on a functional manifold. To that matter, we introduce a new Functional Instance Normalization layer and residual mechanism, which together disentangle image content from position on target manifold. We rely on naive physics-inspired models to guide the training while allowing private model/translations features. CoMoGAN can be used with any GAN backbone and allows new types of image translation, such as cyclic image translation like timelapse generation, or detached linear translation. On all datasets, it outperforms the literature. Our code is available at http://github.com/cv-rits/CoMoGAN .Keywords
All Related Versions
This publication has 55 references indexed in Scilit:
- Continuous and Diverse Image-to-Image Translation via Signed Attribute VectorsInternational Journal of Computer Vision, 2022
- COCO-FUNIT: Few-Shot Unsupervised Image Translation with a Content Conditioned Style Encoder2020
- TSIT: A Simple and Versatile Framework for Image-to-Image Translation2020
- ForkGAN: Seeing into the Rainy Night2020
- Model-Based Occlusion Disentanglement for Image-to-Image Translation2020
- Contrastive Learning for Unpaired Image-to-Image Translation2020
- Rain Rendering for Evaluating and Improving Robustness to Bad WeatherInternational Journal of Computer Vision, 2020
- Multimodal Unsupervised Image-to-Image Translation2018
- Diverse Image-to-Image Translation via Disentangled Representations2018
- Semantic Foggy Scene Understanding with Synthetic DataInternational Journal of Computer Vision, 2018