Few-shot Image Generation via Cross-domain Correspondence
arXiv:2104.06820
Abstract
Training generative models, such as GANs, on a target domain containing limited examples (e.g., 10) can easily result in overfitting. In this work, we seek to utilize a large source domain for pretraining and transfer the diversity information from source to target. We propose to preserve the relative similarities and differences between instances in the source via a novel cross-domain distance consistency loss. To further reduce overfitting, we present an anchor-based strategy to encourage different levels of realism over different regions in the latent space. With extensive results in both photorealistic and non-photorealistic domains, we demonstrate qualitatively and quantitatively that our few-shot model automatically discovers correspondences between source and target domains and generates more diverse and realistic images than previous methods.
arXiv commentCVPR 2021
References in corpus (13)
- Overcoming catastrophic forgetting in neural networks
- A Simple Framework for Contrastive Learning of Visual Representations
- Prototypical Networks for Few-shot Learning
- GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium
- Large Scale GAN Training for High Fidelity Natural Image Synthesis
- LSUN: Construction of a Large-scale Image Dataset using Deep Learning with Humans in the Loop
- Texture Networks: Feed-forward Synthesis of Textures and Stylized Images
- On First-Order Meta-Learning Algorithms
- Diversity-Sensitive Conditional Generative Adversarial Networks
- Freeze the Discriminator: a Simple Baseline for Fine-Tuning GANs
- Few-Shot Unsupervised Image-to-Image Translation
- Few-shot Image Generation with Elastic Weight Consolidation
- Transferring GANs: generating images from limited data