ofiqpy — a faithful Python port of OFIQ v1.1.0
==============================================

ofiqpy reproduces the algorithms of OFIQ (Open Source Face Image Quality), the
ISO/IEC 29794-5 reference implementation developed by the German Federal Office
for Information Security (BSI), Copyright (c) 2024, MIT-licensed:

    https://github.com/BSI-OFIQ/OFIQ-Project

Models are NOT bundled
----------------------
ofiqpy loads OFIQ's own model files at runtime (SSD face detector, ADNet-98
landmarks, 3DDFA-V2 pose, BiSeNet face parsing, face-occlusion segmentation,
sharpness random forest, compression CNN, HSEmotion + AdaBoost expression, and
the MagFace unified-quality model). These files ship with OFIQ and may be
licensed separately from the OFIQ source (see OFIQ's LICENSE.md). They are NOT
redistributed with ofiqpy.

To run ofiqpy, obtain the models from an OFIQ install and set:

    export OFIQPY_OFIQ_DATA=/path/to/OFIQ-Project/data

Conformance
-----------
ofiqpy targets agreement with the OFIQ reference to within +/- 1 quality point
per component (the ISO/IEC 29794-5 Annex A criterion). See the docs for the
validated conformance figures.
