Depth Any Camera: Zero-Shot Metric Depth Estimation from Any Camera
Yuliang Guo, Sparsh Garg, S. Mahdi H. Miangoleh, Xinyu Huang, Liu Ren · 2025
Read Depth Any Camera: Zero-Shot Metric Depth Estimation from Any Camera by Yuliang Guo, Sparsh Garg, S. Mahdi H. Miangoleh, Xinyu Huang, Liu Ren (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-02464.txt.
Description
Yuliang Guo and co-authors teach a depth-estimating model trained on ordinary photos to judge distance through fisheye and full 360-degree cameras. Such wide views bend space severely, and earlier depth models, however good on normal images, fumble them. Their system reframes every image on a standard wide projection, so one representation serves all camera types.
Training tricks do the rest: the model learns from tilted slices that imitate distorted views, aligns training across field-of-view settings, and practices at several resolutions. It never sees a real fisheye photo in training, yet on wide-view test sets it raised the standard accuracy measure by up to fifty percent over rival depth models. The authors argue that with the right representation, the mass of ordinary three-dimensional training data can serve cameras it was never gathered from.
Catalog record
| CREATOR | Yuliang Guo, Sparsh Garg, S. Mahdi H. Miangoleh, Xinyu Huang, Liu Ren |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 63 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-02464