FIGR: Few-shot Image Generation with Reptile
Louis Clouâtre, Marc Demers
TL;DR
This work introduces FIGR, a framework that meta-trains Generative Adversarial Networks using the Reptile algorithm to achieve few-shot image generation. By training on multiple related classes, FIGR enables rapid adaptation to unseen classes using very small samples, demonstrated on MNIST and Omniglot, and extended with the large FIGR-8 dataset. The approach eliminates the need for lengthy inference or external memory and shows promise for generating novel concepts from minimal data. The FIGR-8 benchmark provides a challenging testbed for future meta-learning in GANs and paves the way for broader applications in creative and transfer learning scenarios.
Abstract
Generative Adversarial Networks (GAN) boast impressive capacity to generate realistic images. However, like much of the field of deep learning, they require an inordinate amount of data to produce results, thereby limiting their usefulness in generating novelty. In the same vein, recent advances in meta-learning have opened the door to many few-shot learning applications. In the present work, we propose Few-shot Image Generation using Reptile (FIGR), a GAN meta-trained with Reptile. Our model successfully generates novel images on both MNIST and Omniglot with as little as 4 images from an unseen class. We further contribute FIGR-8, a new dataset for few-shot image generation, which contains 1,548,944 icons categorized in over 18,409 classes. Trained on FIGR-8, initial results show that our model can generalize to more advanced concepts (such as "bird" and "knife") from as few as 8 samples from a previously unseen class of images and as little as 10 training steps through those 8 images. This work demonstrates the potential of training a GAN for few-shot image generation and aims to set a new benchmark for future work in the domain.