DepthAnything/Video-Depth-Anything
Video depth estimation model that produces consistent depth maps across arbitrarily long videos using transformer architecture.
Not currently ranked — collecting fresh signals.
star history
Video Depth Anything extends Depth Anything V2 to handle video sequences, enabling consistent depth estimation across long videos without compromising quality or generalization. It uses a transformer-based architecture and supports both relative and metric depth estimation modes, including streaming inference for real-time applications. The model offers faster inference and fewer parameters compared to diffusion-based alternatives while maintaining higher accuracy.
Frequently asked
- What is DepthAnything/Video-Depth-Anything?
- Video depth estimation model that produces consistent depth maps across arbitrarily long videos using transformer architecture.
- Is Video-Depth-Anything open source?
- Yes — DepthAnything/Video-Depth-Anything is open source, released under the Apache-2.0 license.
- What language is Video-Depth-Anything written in?
- DepthAnything/Video-Depth-Anything is primarily written in Python.
- How popular is Video-Depth-Anything?
- DepthAnything/Video-Depth-Anything has 2k stars on GitHub.
- Where can I find Video-Depth-Anything?
- DepthAnything/Video-Depth-Anything is on GitHub at https://github.com/DepthAnything/Video-Depth-Anything.