视频生成扩散模型综述：架构、采样与挑战

AI 摘要

本文介绍了扩散模型在视频生成领域的最新进展，包括从零训练视频扩散模型的方法、参数化与采样基础、3D U-Net和Transformer架构，以及Google和OpenAI等机构的相关工作。文章强调了视频生成相比图像生成的额外挑战，如时间一致性和数据稀缺性，并讨论了重建引导和v-参数化等技术。

正文节选

Diffusion models have demonstrated strong results on image synthesis in past years. Now the research community has started working on a harder task—using it for video generation. The task itself is a superset of the image case, since an image is a video of 1 frame, and it is much more challenging because: - It has extra requirements on temporal consistency across frames in time, which naturally demands more world knowledge to be encoded into the model. - In comparison to text or images, it is mo

发布时间：2024-04-12 08:00

抓取时间：2026-08-02 00:26

来源机构：Lilian Weng