HTTP/2 200 server: GitHub.com content-type: text/html; charset=utf-8 last-modified: Thu, 27 Feb 2025 07:59:59 GMT access-control-allow-origin: * strict-transport-security: max-age=31556952 etag: W/"67c01b7f-7b0f" expires: Tue, 30 Dec 2025 04:19:49 GMT cache-control: max-age=600 content-encoding: gzip x-proxy-cache: MISS x-github-request-id: 6167:2D64E0:984F4E:AB378B:6953508A accept-ranges: bytes age: 0 date: Tue, 30 Dec 2025 04:09:49 GMT via: 1.1 varnish x-served-by: cache-bom-vanm7210074-BOM x-cache: MISS x-cache-hits: 0 x-timer: S1767067790.722533,VS0,VE207 vary: Accept-Encoding x-fastly-request-id: b4899d55ac100a6fe316145c3584f342e0028d19 content-length: 3743

Video Depth Anything: Consistent Depth Estimation for Super-Long Videos

Sili Chen      Hengkai Guo^†      Shengnan Zhu      Feihu Zhang     
Zilong Huang      Jiashi Feng      Bingyi Kang^†     
ByteDance
†Corresponding Author

CVPR 2025

arXiv Code 🤗 Hugging Face Demo

This work presents Video Depth Anything, which is based on Depth Anything V2. It can be applied to arbitrarily long videos and exhibits strong quality, temporal consistency, and generalization ability.

Long Video Results

Comparison with DepthCrafter on open-world videos

Comparison with DepthCrafter on Unprojected Point Cloud

Intrinsic of camera as well as the aligned scale and shift parameters are obtained by running the first frame of the video through MoGe, which is a powerful model for recovering 3D geometry from monocular open-domain images.

Citation

@article{video_depth_anything,
    title={Video Depth Anything: Consistent Depth Estimation for Super-Long Videos},
    author={Chen, Sili and Guo, Hengkai and Zhu, Shengnan and Zhang, Feihu and Huang, Zilong and Feng, Jiashi and Kang, Bingyi},
    journal={arXiv:2501.12375},
    year={2025}
  }