Hierarchical Patch VAE-GAN: Generating Diverse Videos from a Single Sample

We consider the task of generating diverse and novel videos from a single\nvideo sample. Recently, new hierarchical patch-GAN based approaches were\nproposed for generating diverse images, given only a single sample at training\ntime. Moving to videos, these approaches fail to generate diverse samples, and\noften collapse into generating samples similar to the training video. We\nintroduce a novel patch-based variational autoencoder (VAE) which allows for a\nmuch greater diversity in generation. Using this tool, a new hierarchical video\ngeneration scheme is constructed: at coarse scales, our patch-VAE is employed,\nensuring samples are of high diversity. Subsequently, at finer scales, a\npatch-GAN renders the fine details, resulting in high quality videos. Our\nexperiments show that the proposed method produces diverse samples in both the\nimage domain, and the more challenging video domain.\n

Paper

Similar papers

© 2026 NYSGPT2525 LLC