CoMoGAN: continuous model-guided image-to-image translation
Authors:
- 1 June 2021
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 14283-14293
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 in this page: https://github.com/cv-rits/CoMoGAN.Keywords
All Related Versions
This publication has 49 references indexed in Scilit:
- Unpaired Image-to-Image Translation Using Cycle-Consistent Adversarial Networks2017
- Deep Feature Interpolation for Image Content Changes2017
- Improved Texture Networks: Maximizing Quality and Diversity in Feed-Forward Stylization and Texture Synthesis2017
- Pyramid Scene Parsing Network2017
- Image-to-Image Translation with Conditional Adversarial Networks2017
- The SYNTHIA Dataset: A Large Collection of Synthetic Images for Semantic Segmentation of Urban Scenes2016
- Deep Residual Learning for Image Recognition2016
- The Cityscapes Dataset for Semantic Urban Scene Understanding2016
- An analytic model for full spectral sky-dome radianceACM Transactions on Graphics, 2012
- A Spatial Post-Processing Algorithm for Images of Night ScenesJournal of Graphics Tools, 2002