[general]
name=WZ Workflow
qgisMinimumVersion=3.0
qgisMaximumVersion=3.99
description=Comprehensive tool for Polish urban planning analysis (Building Permit Conditions - Warunki Zabudowy)
version=2.0.3
author=Adrian Linkowski
email=link.mapy@gmail.com

about=Advanced QGIS plugin for automated urban planning analysis in Poland (Warunki Zabudowy - WZ).

    Features:
    • 14-step automated workflow for building permit analysis
    • Deep-learning (CNN) roof-shape classification
    • Random Forest classification of biologically active surface (PBC) from LiDAR
    • LiDAR point cloud processing, robust on very large clouds
    • Cadastral data integration (ULDK/BDOT)
    • Automated Word report generation
    • Urban planning indicators (WPZ, WIZ, WPBC)

    ML models (~218 MB) are downloaded automatically on first use from GitHub Releases.

    ---

    Kompleksowe narzędzie do automatycznej analizy warunków zabudowy w Polsce.
    Automatyzuje 14-stopniowy proces analizy urbanistycznej z wykorzystaniem uczenia maszynowego
    (klasyfikacja kształtu dachów siecią CNN, klasyfikacja powierzchni biologicznie czynnej
    lasem losowym na podstawie danych LiDAR).
    Modele ML (~218 MB) są pobierane automatycznie przy pierwszym uruchomieniu z GitHub Releases.

tracker=https://github.com/AdrLin/wz-workflow/issues
repository=https://github.com/AdrLin/wz-workflow
homepage=https://github.com/AdrLin/wz-workflow

tags=python,urban planning,poland,cadastral,analysis,building,machine learning,deep learning,cnn,random forest,lidar,permits

category=Vector
icon=icon.png

experimental=False
deprecated=False
changelog=2.0.3
    - Removed automatic package installation via shell (os.system pip install); missing dependencies now raise a clear message
    - Documented and hardened the subprocess call used to launch the roof classifier
    - Replaced bare except clauses with explicit exception handling
    2.0.2
    - Updated Word templates
    2.0.1
    - Replaced exec() with the standard runpy mechanism for running workflow steps (security hardening)
    - Model downloader now validates the URL scheme and host, and streams the download (security hardening)
    2.0.0
    - Completely rebuilt PBC (biologically active surface) classifier: Random Forest replaces the previous SVM/MLP approach
    - New 13-class surface taxonomy reflecting real-world land cover encountered in WZ analyses
    - 28 engineered features including chromaticity, spectral ratio indices and neighbourhood geometry
    - Configurable decision threshold lets the model err on the conservative side when uncertain
    - New deep-learning (CNN, EfficientNetV2) model for roof-shape classification, replacing the earlier classifier
    - Improved robustness for extremely large LiDAR point clouds (memory-safe per-parcel processing)
    - Prediction layer now stores per-hexagon class probability for quality inspection
    - New plugin icon; several internal scripts refined and stabilised
    1.0.4
    - The script for determining the analyzed area and the descriptive analysis templates were adapted to reflect the latest changes in the law
    1.0.3
    - Minor fixes and template updates
    1.0.2
    - Fixed plugin packaging (removed large files from ZIP)
    1.0.1
    - Added automatic municipality name extraction from granica_terenu layer
    - {{nazwa_gminy}} placeholder available in Word templates
    - Updated Word report templates
    1.0.0
    - Initial public release
    - Full 14-step WZ analysis workflow
    - Automatic ML model download
    - Roof-shape classification (flat/gable/hip/shed)
    - Point cloud processing (LiDAR)
    - Automated Word report generation
    - Urban planning indicators calculation

hasProcessingProvider=no
