Real-world scale estimation using agisoft software

[pjdeemac]

Hi!

Is there a way for Agisoft Metashape or Meshroom to automatically recognize the real-world scale of a scene,
based only on geometric information in the images - without placing any reference object (like a ruler or marker)?

In other words: can metashape infer actual size from visual clues alone, or is a known dimension always required?

Can I do so importing camera parameters as focal length and sensor width?

Thanks!

[OBNRacerMan]

Conventional photogrammetry is not calibrated against any standard, as is the case with 3D scanners, for this reason it is simply impossible to calculate the scale of the received content to any denominator, except for using known data on the sizes of some details in the image (that is why coded marks with known distances between them are used). P.S. I have a Revopoint Miraco Plus 3D scanner, it comes with coded rulers and markers for scanning large objects with high accuracy, combining photogrammetry and "pure" 3D scanning modes.

[Micr0]

Hi!

Is there a way for Agisoft Metashape or Meshroom to automatically recognize the real-world scale of a scene,
based only on geometric information in the images - without placing any reference object (like a ruler or marker)?

In other words: can metashape infer actual size from visual clues alone, or is a known dimension always required?

Can I do so importing camera parameters as focal length and sensor width?

Thanks!

A while back, I worked with a software engineer who was creating an AI based app that could that could map the three dimensional depths in a scan to a gray scale image that was read by a panto graph type machine. It seems an ai could probably Be used to recognize known objects in a 3D scan, and use their dimensions as a reference for the rest of the scan. That, however, is way above my pay grade..

[matthwadams]

No, Agisoft Metashape or Meshroom cannot automatically determine true real-world scale without a known reference dimension. Visual clues alone (like object shapes or perspective) are not enough for photogrammetry software to infer scale with high accuracy.

Here’s why:
Photogrammetry works by calculating relative geometry based on image overlap and parallax. Without any known size (like a reference object or GCP), the output model will be dimensionally accurate in proportion but scale-agnostic — meaning it's the correct shape, but the software doesn't know if it's 1cm or 1m.

That said, importing camera parameters like:
Focal length
Sensor width
Pixel size

does help improve camera alignment and model accuracy, but it still doesn't establish scale. These parameters ensure a more accurate internal camera model, reducing distortion and improving relative measurements — but again, you still need at least one known dimension or scale constraint to convert the model from arbitrary units to real-world units.

Cheer's,
Mattew