# Depth Pro 2: Sharp and Temporally Consistent Metric Monocular Depth

> Source: <https://huggingface.co/papers/2611.26603>
> Published: 2026-11-20 16:25:25+00:00

**Key takeaway:** A single network now predicts metric depth at native resolution for both single images and video, removing the choice between multi-frame temporal consistency and single-frame boundary sharpness.

[Depth Estimation •  Updated     •  76](https://huggingface.co/apple/DepthPro2)  

[Papers](https://huggingface.co/papers)

# Depth Pro 2: Sharp and Temporally Consistent Metric Monocular Depth

## Abstract

A single scalable network estimates dense metric depth for images and video at native resolution, jointly improving metric accuracy, boundary sharpness, and temporal consistency.

[deepseek-ai/DeepSeek-V4.1-Flash](https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash)

We present Depth Pro 2, an approach to estimating [dense metric depth](https://huggingface.co/papers?q=dense%20metric%20depth) for images and video at full resolution. By analyzing prior methods for [monocular depth estimation](https://huggingface.co/papers?q=monocular%20depth%20estimation), we find that they commonly trade off [temporal consistency](https://huggingface.co/papers?q=temporal%20consistency) and [boundary sharpness](https://huggingface.co/papers?q=boundary%20sharpness): [multi-frame methods](https://huggingface.co/papers?q=multi-frame%20methods) favor consistency, [single-frame methods](https://huggingface.co/papers?q=single-frame%20methods) favor sharpness. To overcome this trade-off, we train a single, [scalable network](https://huggingface.co/papers?q=scalable%20network) that processes both images and video and predicts [depth maps](https://huggingface.co/papers?q=depth%20maps) at [native resolution](https://huggingface.co/papers?q=native%20resolution). Depth Pro 2 produces [depth maps](https://huggingface.co/papers?q=depth%20maps) that are metrically accurate, sharp, and temporally consistent all at once, outperforming prior work in all of these dimensions. We release a family of models that define a new [Pareto frontier](https://huggingface.co/papers?q=Pareto%20frontier) in terms of [metric accuracy](https://huggingface.co/papers?q=metric%20accuracy), sharpness, [temporal consistency](https://huggingface.co/papers?q=temporal%20consistency), and runtime. We release code and weights at https://apple-aiml-research.github.io/ml-depth-pro-2/

### Community

Depth Pro 2 estimates the depth of every pixel, in meters, from images taken with an ordinary camera. One network handles a single photo, a whole video, or a live camera feed, at the input's full resolution. Its depth is sharp at object edges and steady from frame to frame.

Get this paper in your agent:

`hf papers read 2611.26603` ## Don't have the latest CLI?

`curl -LsSf https://hf.co/cli/install.sh | bash` ## Models citing this paper 10

    [Depth Estimation •  23.5M • Updated   •  42  •  14](https://huggingface.co/apple/DepthPro2-focal)  

#### apple/DepthPro2-focal

[Browse 10 models citing this paper](https://huggingface.co/models?other=arxiv:2611.26603)

## Datasets citing this paper 0

No dataset linking this paper

### Spaces citing this paper 3

## Collections including this paper 0

No Collection including this paper

[collection](https://huggingface.co/new-collection)to link it from this page.
