Skip to content

system_config.md

Latest commit

 

History

History
executable file
·
79 lines (60 loc) · 3.68 KB

system_config.md

File metadata and controls

executable file
·
79 lines (60 loc) · 3.68 KB
Raw

System Configuration

The system configuration declares the biometric baselines that experiments may select and maps database names to their protocol and raw-data locations. The repository provides an example at config/system_config.yaml.

The example uses /path/to/... placeholders for raw datasets. Replace them with your local paths before running an experiment. Relative paths are resolved from the process's current working directory, not from the directory containing the YAML file; the examples below assume commands are run from the repository root.

Structure

Baselines

baselines:
  - iresnet100
  - iresnet50
  - edgeface
  - edgefacexs
  - facenet
  - random

baselines is a configuration whitelist: an experiment's bio_alg must appear here. Implementations are registered separately by get_baseline_dict() in src/btpbench/algorithms.py, so a baseline must be present in both places to be usable. Adding a name to the YAML does not implement or register it. The random baseline generates random embeddings for testing purposes.

Protected Baselines

protected_baselines:
  - biohash
  - polyprotect
  - combined

This list describes the registered Biometric Template Protection (BTP) algorithms, but is currently informational: the runtime does not read protected_baselines when validating or constructing a BTP algorithm. BTP implementations are selected from get_protected_baseline_dict() in src/btpbench/algorithms.py. Editing this YAML list therefore does not enable or disable an implementation. When adding a BTP implementation, update the code registry and keep this informational list synchronized where appropriate.

Databases

Each database entry contains:

databases:
  soteria:
    proto_dir: ./protocols/soteria
    verification_samples: ./protocols/soteria/soteria-verification.csv
    unlink_samples: ./protocols/soteria/soteria-bf-no-mask-samples.csv
    dataset_dir: /path/to/soteria
    extension: ""
    video: true

The provided configuration contains protocol bundles for soteria, icarb, mobio, and multipie, plus test, a small protocol fixture and configuration example. The corresponding raw media are not distributed with this project.

The fields have the following meanings:

Field Description
proto_dir Root directory containing evaluation and irreversibility protocol CSVs.
verification_samples Sample list used by verification and the default diversity flow.
unlink_samples Sample list used by unlinkability and unlinkability-template diversity runs.
dataset_dir Root directory containing the raw image or video files; these files are not included in this repository.
extension Text appended verbatim to every CSV path; use an empty string when paths already contain their suffix.
video Selects the video loader when true and the image loader when false.

All four path fields are interpreted relative to the current working directory when they are not absolute. Both sample-list CSVs are read when a dataset is initialized, so verification_samples and unlink_samples must both point to readable files even when a particular command uses only one of them.

Protocol files

All protocol CSVs share the required columns path, subject_id, and template_id. Identification uses paired for_enrolling.csv and for_probing.csv files below protocol and split directories, while irreversibility uses two fixed files below irreversibility/.

See Protocol Format and Custom Databases for the complete CSV schema, directory layout, path and video-frame behavior, evaluation-set semantics, metadata constraints, worked example, and validation checklist.