We present Buffer Anytime, a framework for estimation of depth and normal maps (which we call geometric buffers) from video that eliminates the need for paired video--depth and video--normal training data. Instead of relying on large-scale annotated video datasets, we demonstrate high-quality video buffer estimation by leveraging single-image priors with temporal consistency constraints. Our zero-shot training strategy combines state-of-the-art image estimation models based on optical flow smoothness through a hybrid loss function, implemented via a lightweight temporal attention architecture. Applied to leading image models like Depth Anything V2 and Marigold-E2E-FT, our approach significantly improves temporal consistency while maintaining accuracy. Experiments show that our method not only outperforms image-based approaches but also achieves results comparable to state-of-the-art video models trained on large-scale paired video datasets, despite using no such paired video data.
@article{arxiv.2411.17249,
title = {Buffer Anytime: Zero-Shot Video Depth and Normal from Image Priors},
author = {Zhengfei Kuang and Tianyuan Zhang and Kai Zhang and Hao Tan and Sai Bi and Yiwei Hu and Zexiang Xu and Milos Hasan and Gordon Wetzstein and Fujun Luan},
journal= {arXiv preprint arXiv:2411.17249},
year = {2024}
}