KG-GAN: Knowledge-Guided Generative Adversarial Networks
arXiv:1905.12261
Abstract
Can generative adversarial networks (GANs) generate roses of various colors given only roses of red petals as input? The answer is negative, since GANs' discriminator would reject all roses of unseen petal colors. In this study, we propose knowledge-guided GAN (KG-GAN) to fuse domain knowledge with the GAN framework. KG-GAN trains two generators; one learns from data whereas the other learns from knowledge with a constraint function. Experimental results demonstrate the effectiveness of KG-GAN in generating unseen flower categories from seen categories given textual descriptions of the unseen ones.
References in corpus (12)
- Conditional Generative Adversarial Nets
- Spectral Normalization for Generative Adversarial Networks
- Generative Adversarial Text to Image Synthesis
- Improved Training of Wasserstein GANs
- Semantic Image Synthesis with Spatially-Adaptive Normalization
- CAN: Creative Adversarial Networks, Generating "Art" by Learning About Styles and Deviating from Style Norms
- Learning Deep Representations of Fine-grained Visual Descriptions
- Zero-Shot Learning -- The Good, the Bad and the Ugly
- Deep Generative Models with Learnable Knowledge Constraints
- Mode Seeking Generative Adversarial Networks for Diverse Image Synthesis
- Modular Generative Adversarial Networks
- IAN: Combining Generative Adversarial Networks for Imaginative Face Generation