{"name": "QMaxent", "package_name": "qmaxent", "version": "0.2.0", "experimental": false, "qgis_min": "3.44.0", "qgis_max": "3.99.0", "downloads": 19, "uploaded_by": "bhyu", "upload_datetime": "2026-10-29T12:19:33.575457", "changelog": "0.2.0\nInstallation and robustness\n- Fix: dependency installation failed on QGIS 3.44.15+ for Windows\n(\"ensurepip ... returned non-zero exit status 1\"). Packages are now\ninstalled with the QGIS Python's own pip into the QMaxent package\nfolder (no venv interpreter needed on Windows).\n- Fix: Harmonize stopped with a raw PROJ error for rasters whose CRS\nis a local/engineering system without a datum. Check Raster\nConsistency now names the raster, explains the problem, suggests\nthe matching EPSG code and how to fix it; a CRS set in QGIS\n(Layer CRS > Set Layer CRS...) is used for such rasters.\n- Fix: the Data-tab raster list now follows the QGIS project (no\nstale rows after a new project or removed layers); missing inputs\nare reported in a dialog instead of the Training log.\n- Checks before training: multi-band rasters are rejected with\ninstructions; the log reports presences dropped (outside rasters\nor NoData) and presences sharing a raster cell.\nInputs and outputs\n- New: Open point file... (any QGIS point format) and Load CSV (X/Y\ncolumns)... with delimiter, column and CRS detection.\n- New: Add raster files... (any GDAL single-band raster).\n- New: <model>_model_card.json (settings, data, performance,\nthresholds, variable importance, software versions) and\n<model>_coefficients.csv (all Maxent features and coefficients)\nare written next to model.pkl.\nMethods and wording\n- Permutation importance: choose the evaluation data, Training data\n(maxent.jar convention, default) or Held-out CV test folds (all\nfolds pooled; v0.1.x used the first fold only).\n- Auto feature types follow the maxnet default rule (Phillips et al.\n2017): 80 or more presences now give LQPH instead of LQPHT, because\nthreshold features overfit with the elapid scikit-learn backend\n(Bradypus random 10-fold test AUC 0.873 with T, 0.890 without).\nThreshold features remain available under Manual selection.\n- Fix: projection no longer fails when a categorical raster contains\na class that was absent from the training data.\n- Fix: running survey site selection again no longer fails when the\noutput GeoPackage already exists (it is replaced, or a new file\nname is used if another program has it open).\n- Survey tab renamed Survey Site Selection; modes renamed\nHigh-suitability search and Gradient-stratified sampling.\n- Reverse geocoding works worldwide with a choice of address\nlanguage (local names, QGIS language, English).\n- Shorter CRS labels in the status line and training log.\n- Fix random seed now also covers background sampling (v0.1.x drew a\ndifferent background on every run).", "external_deps": null, "download_url": "https://plugins.qgis.org/plugins/qmaxent/version/0.2.0/download/"}