Configuration Files

VISWIR vQuasar relies on a set of configuration files to control input/output paths, fusion parameters, and optional detection settings. This section explains the purpose of each file and provides examples.

Base Configuration

  • base_config.yaml : defines input/output folders, reference image, mode, etc.

  • parameters.json : defines fusion parameters.

  • yolo_config.json : YOLO detection parameters used if run_detection: true in base_config.yaml (optional, defaults are used if missing).

Note

The base configuration is required for most runs. It specifies where data is located and how the fusion pipeline should behave.

Fast Pipeline Configuration

  • fast_config.yaml : minimal config for the fast pipeline.

  • yolo_config.json : YOLO detection parameters (optional, defaults are used if missing).

Important

The fast mode skips metric computation. It is intended for rapid prototyping and quick previews, not for final evaluation.

Examples

Example of a minimal base configuration:

Example base_config.yaml
# Paramètres de pilotage
visible_folder: "../data/VIS"
swir_folder: "../data/SWIR"
output_folder: "../results" # sortie
ref_image_path: null # pour le calcul des métriques R-IQA, nécessite une image de référence par paire de fusion
ground_truth_path: "../data/Ground truth"

mode: "sql"   # "fixed", "sql", "optuna"

run_detection: false # true, false
save_output: true # enregistre ou non les images fusionnées

# Ici on ne met PAS les paramètres de fusion (ils restent dans parameters.json)

Example of fusion parameters (JSON):

{
   "mode_fixe": true,
   "facteur_swir": 0.89,
   "beta": 1.07,
   "level": 5,
   "apply_gamma": true,
   "gamma_value": 2.82
}

Example of YOLO detection config:

{
   "model_path": "yolov8x.pt",
   "confidence_threshold": 0.25,
   "iou_threshold": 0.3,
   "device": "cpu",
   "save_detection_results": false,
   "allowed_classes": [
      "truck",
      "person",
      "bus",
      "motorcycle",
      "bicycle",
      "car"
   ]
}