Set up the data and the hardware
The package installs as the Installation says, with every dependency from the General registry. This page is for what the package does not install: the hardware a calibration runs on, and the data of the experiments.
Clone
The repository holds the package (RadiometricCamera/), the experiments that reproduce the paper (experiments/), the manuscript and the project page, and no data:
git clone https://github.com/RealCamerasAreRadiometric/realcamerasareradiometric.github.io.git
cd realcamerasareradiometric.github.io
julia +1.13 --project=RadiometricCamera -e 'using Pkg; Pkg.instantiate()'What runs where
| piece | needed for |
|---|---|
| Julia 1.13 | everything |
| a CUDA GPU with ≥ 20 GiB free (optional) | the :gpu test items (Run the tests) and fast refinements; the CPU backend runs everything, more slowly (JULIA_NUM_THREADS=auto: it is multithreaded over pixels) |
| a calibration session (VisualGeometryToolkitIO records and the photographs) | Calibrate a camera on one's own camera |
| the published data of the experiments | the paper's experiments, the example walkthrough examples/radiometric_calibration.jl and the end-to-end checks of Check a change; not the package tests |
The package tests need no data: their fixtures are committed under test/fixtures/.
Data
The experiments' inputs — the raw photographs, the synthetic renders, the baseline calibrations of the external front ends (Schöps et al.'s pipeline, Kalibr, OpenCV, BlobBoards), the targets and the calibration cases — are the paper project's published data, tiers of lazy Julia artifacts catalogued in experiments/configs/data_artifacts.toml. experiments/REPRODUCE.md says how to fetch them (experiments/scripts/fetch_data.jl) and how to bind a copy of one's own (VGTK_DATA_ROOT); the package's checks name the tier they read (rc_calibration).