Contrastive Learning for Unpaired Image-to-Image Translation
arXiv:2007.15651
Abstract
In image-to-image translation, each patch in the output should reflect the content of the corresponding patch in the input, independent of domain. We propose a straightforward method for doing so -- maximizing mutual information between the two, using a framework based on contrastive learning. The method encourages two elements (corresponding patches) to map to a similar point in a learned feature space, relative to other elements (other patches) in the dataset, referred to as negatives. We explore several critical design choices for making contrastive learning effective in the image synthesis setting. Notably, we use a multilayer, patch-based approach, rather than operate on entire images. Furthermore, we draw negatives from within the input image itself, rather than from the rest of the dataset. We demonstrate that our framework enables one-sided translation in the unpaired image-to-image translation setting, while improving quality and reducing training time. In addition, our method can even be extended to the training setting where each "domain" is only a single image.
ECCV 2020. Please visit https://taesungp.github.io/ContrastiveUnpairedTranslation/ for introduction videos and more. v3 contains typo fixes and citation update
References in corpus (2)
Cited by in corpus (6)
- Contrastive Representation Learning: A Framework and Review
- Contrastive Learning for Compact Single Image Dehazing
- Contrastive Unpaired Translation using Focal Loss for Patch Classification
- Visual Diver Face Recognition for Underwater Human-Robot Interaction
- Deep Learning-based Frozen Section to FFPE Translation
- Synthesizing Photorealistic Images with Deep Generative Learning