Summary of Buffer Anytime: Zero-shot Video Depth and Normal From Image Priors, by Zhengfei Kuang et al.
Buffer Anytime: Zero-Shot Video Depth and Normal from Image Priors
by Zhengfei Kuang, Tianyuan Zhang, Kai Zhang, Hao Tan, Sai Bi, Yiwei Hu, Zexiang Xu, Milos Hasan, Gordon Wetzstein, Fujun Luan
First submitted to arxiv on: 26 Nov 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The Buffer Anytime framework eliminates the need for paired video-depth and video-normal training data by leveraging single-image priors with temporal consistency constraints. It 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, the approach significantly improves temporal consistency while maintaining accuracy. The method outperforms image-based approaches and achieves results comparable to state-of-the-art video models trained on large-scale paired video datasets, despite using no such paired video data. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Buffer Anytime is a new way to estimate depth and normal maps from videos without needing lots of labeled training data. Instead, it uses information from single images and makes sure the results make sense over time. This approach works well with popular image estimation models and can even match the quality of more complex video-based methods that require much more training data. |
Keywords
» Artificial intelligence » Attention » Loss function » Optical flow
