Skip to content
imaged

Metric depth

Depth Anything V3 metric (Large)

Estimate per-pixel depth in meters from a single photo.

Apache-2.0
Explore model

About this model

Depth Anything V3 metric is a ViT-L backbone fine-tuned to return absolute depth in meters, not relative disparity. Fixed input is 280x504. The build is a 697 MB fp32 ONNX from TillBeemelmanns on HuggingFace.

On M1 Max with CoreML, warm inference runs in 265 ms, about 7 times faster than CPU. Metric values are most reliable when the scene resembles the training distribution; otherwise treat depth as relative within one frame.

What you can build

  • Estimate distance to objects in a single photo for AR previews.
  • Generate depth-aware blur or bokeh in a photo editor.
  • Build a rough 3D scene reconstruction from a single frame.
  • Pre-compute depth for downstream point-cloud or mesh tools.

Variants & downloads

Choose the build for your app

Model files are being prepared.

Explore the model’s examples, source links, and integration notes while download variants are added.

Example inference

Input
A close-up cat photograph from Wikimedia Commons.
Output
Side-by-side input and an INFERNO-colored depth map of the cat scene.

Performance

Results depend on the device, runtime, and test conditions. Each measurement keeps its own context.

M1 Max, macOS 15, ONNX Runtime 1.17

See original catalog hardware notes · CPU

1.83 s

Warm inference

Catalog reference · variant unspecifiedCold start (ms): 1831.13

Reference result from the existing catalog; conversion variant and measurement date were not recorded.

M1 Max, macOS 15, ONNX Runtime 1.17

See original catalog hardware notes · CoreML

265.0 ms

Warm inference

Catalog reference · variant unspecifiedCold start (ms): 1100.29

Reference result from the existing catalog; conversion variant and measurement date were not recorded.

RTX 2060, Ubuntu 24.04, ONNX Runtime 1.24.3

See original catalog hardware notes · CUDA

139.3 ms

Warm inference

Catalog reference · variant unspecified

Reference result from the existing catalog; conversion variant and measurement date were not recorded.

Integration examples

Vision SDK integration

cpp

#include "imaged/sdk.hpp"
#include "imaged/families/depth/depth_family.hpp"

imaged::SDK::AI ai;
auto& config = ai.getOptions();
config.set("depth.model", "weights/depth_anything_v3_metric_large.onnx");
config.set("depth.inference.provider", "coreml");

imaged::SDK::DepthFamily family;
family.load(config);

imaged::SDK::Image image;
image.load("scene.jpg");

auto result = family.run(&image);
auto* depth = std::get_if<imaged::SDK::DepthResult>(&result);
// depth->depth holds 280*504 floats in meters