Fiducia

The free, open source photogrammetry workhorse

Welcome to Fiducia. It is a photogrammetry workstation built to replace old, crummy software that has a tendency to crash 30 minutes before a looming deadline. If that's the situation you find yourself in, then there is good news. What would have taken hours previously, you can complete in under 30 minutes. Pause, get a cup of coffee, and give Fiducia a try.

fiducial marks measured on every frame principal point the true image centre drag a corner mark to move the principal point
The Fiducia workspace: a solved bundle adjustment of six aerial photographs, with control and tie points on the photo and the accuracy statement in the panel

A block of six photographs over Lyons, Colorado, solved in Fiducia.

Why Fiducia

Three core pillars define Fiducia

A photogrammetric workflow usually means a lot of windows, a lot of waiting and not much explanation when something goes wrong. These are the three things Fiducia is built around.

perspective centre solved for every image RMS 2.41 m RMS 0.18 m · converged drag a control point off its mark to make a blunder
01 / 03

Transparent

Any outliers are clearly labelled.

  • Residuals appear from the fourth control point, while you're still measuring
  • Blunders are found by data snooping and named, with the reason
  • Every result comes with its standard deviation
orthophoto terrain displacement removed mosaic seamless across the cutline contours and terrain models
02 / 03

Efficient

Highly optimised, logical workflow.

  • Calibration certificates read in seconds, offline
  • Ground control found automatically against a reference orthophoto
  • Four large-format frames from raw photos to orthos, a mosaic and a merged DEM in about ten minutes of processing
snapshot · before re-solving restore to any earlier state, any time
03 / 03

Robust

Every edit is autosaved the moment you make it.

  • Pull out the USB stick mid-edit and Fiducia says so, then carries on when it's back
  • Long jobs run in the background and can be cancelled
  • Memory is watched, so a big orthophoto doesn't take the computer down with it

A closer look

Preview Fiducia

A one canvas interface? Yes please!

The Images step: six photographs listed, one open on the canvas The Control step: ground control and tie points on a photograph, with live residuals The Model step: a converged bundle adjustment with its accuracy statement The Ortho step: an orthophoto open on the canvas The Mosaic step: the live preview coloured by which photo each part comes from The finished mosaic of the six photographs The Stereo DEM step: a stereo pair side by side, with pairs ranked by base-to-height ratio The Terrain step: a shaded relief open on the canvas The command palette open over the workspace


az 0° · el 30° matched 100%
Stereo DEM

Illustration of some behind the scenes magic

This is roughly what happens when you press Extract on the Stereo DEM page. P.S. you can drag the 3D model around

  1. 01

    Match

    Both photos are resampled so the same ground sits on the same row, then every pixel is searched for along that row in its partner by semi-global matching.

  2. 02

    Triangulate

    Each match becomes a pair of rays from the adjusted cameras, intersected in 3D.

  3. 03

    Merge

    Surfaces from every pair are levelled against each other and merged into one, ready for contours, hillshades and volumes.


Accuracy

Science first.

The primary goal of Fiducia is accuracy, because what use is a science software that is inaccurate? The software was built around algorithmic accuracy and has been thoroughly tested with both synthetic and real-world data. This ensures margins of error fall well within all accepted and expected limits. See key testing results below.

tie point accuracy
0.04 px
variance factor σ₀
1.00
lower error with self-calibration
16×
injected blunders caught
2 / 2
GNSS offset recovered
0.1 m
confidence, ASPRS · NSSDA
95%
See the full test results →

Features

Robust processing, Robust features

Camera

Certificates read offline

Windows' built-in OCR and a set of rules I've put together read calibration certificates straight away. Claude or ChatGPT can read the harder ones with your own API key.

Control

Automatic ground control

Photos are placed on a reference orthophoto and control points matched automatically, then measured on every photo that sees them.

Matching

Least-squares tie points

Found on reduced images, refined at full resolution, and linked into multi-ray points that tie a strip together.

Adjustment

A bundle you can trust

Separate weights for every kind of observation, self-calibration, GNSS/IMU positions, robust estimation and data snooping.

Products

Orthos and mosaics

Rectified on one pixel grid, then balanced, dodged and joined along seamlines that avoid buildings.

Terrain

Stereo DEMs

Semi-global matching at the detail you choose, triangulated properly in 3D and merged into one levelled surface.

Measuring

Made for clicking points

A loupe at 1:1 while you measure, the pending point shown before you accept it, and stereo pairs side by side with a Line up button.

Screens

Multi-monitor

Pop any photo, mosaic or DEM out into its own window and measure on the second screen.

Privacy

Your data stays yours

Everything lives on your own computer. Nothing is uploaded unless you choose to use an online reader for a certificate.


Under the hood

Built from first principles

The photogrammetric models (collinearity, resection, bundle adjustment and orthorectification) are my own implementation from first principles. For the numerical optimiser, dense stereo matching, feature matching and raster and projection handling I use established open-source libraries.

Fiducia is open source too, under the Apache 2.0 licence, so anyone can read the maths, check it and build on it.

SciPyOpenCVGDALPROJNumPyrasterio
Where it's at

Very much still a prototype

  • The ortho and mosaic side is looking really promising, and I'm very excited by the speed and intuitiveness of the workflow.
  • The LiDAR side is the earliest part, and satellite imagery only goes as far as refining RPCs for now.
  • Windows only. The installer isn't signed yet, so Windows will warn you before it runs.
  • Next on the list is structure from motion, so UAV and oblique photo sets can be oriented without a camera certificate.

Notes from the creator

Notes

Updates, what I'm working on, and the odd bit of behind the scenes photogrammetry.

All notes →


Download

Download free, or view source code

Fiducia is open source, which means it lives and breathes on community input. Send feature requests, bugs or design improvements through my LinkedIn, or open a pull request on GitHub.

  • Windows 10 or 11, 64-bit
  • Installs just for you, so it works on university and work computers without admin rights
  • The installer isn't signed yet, so Windows SmartScreen will ask first. Choose More info, then Run anyway
Fiducia 0.1.0
Fiducia-Setup-0.1.0.exe
Version
0.1.0
Size
189 MB
Released
6 Oct 2026
Download for Windows View the source on GitHub
SHA-256, to check the download
17c491a10b34f2792e1b16090959122c67a4e38e3950b56960333c90aa87535d
Get in touch

I'd love to see where this can go.

Any feedback or questions are much appreciated, whether something breaks, feels awkward, or you'd expect it to work differently. The easiest way to reach me is on LinkedIn.