Dependencies#

The Geo-SAM plugin requires several Python packages to run. The Dependencies tab in Settings manages their installation inside an isolated, plugin-private directory — no manual pip commands needed.

Installing Dependencies#

Open Geo-SAM Settings and go to the Dependencies tab.
Review the dependency table — missing packages are highlighted in red.
Click Install Missing to begin installation.
Monitor progress in the log area at the bottom of the dialog.
After installation completes, restart QGIS for the changes to take effect.

Note

The first installation may take several minutes because PyTorch is a large package (several hundred MB). Subsequent launches use the cached installation and load much faster.

Required Packages#

Package	Purpose
`torch`	PyTorch deep-learning framework
`torchvision`	Image transforms and utilities for PyTorch
`ultralytics`	SAM model loading and inference
`rasterio`	Reading and writing geospatial raster data
`geopandas`	Vector data handling (Shapefiles)
`shapely`	Geometric operations
`pyproj`	Coordinate reference system transformations
`geosam`	Geo-SAM core library (image encoding and segmentation)

Dependency Status#

Each package shows one of the following statuses in the table:

Installed (runtime) – provided by the QGIS Python environment itself (e.g., rasterio, geopandas).
Installed (plugin-managed) – installed by Geo-SAM into its private dependency directory under .deps/.
Missing (installable) – not yet installed, but the plugin knows how to install it.

Managing Installed Dependencies#

Refresh Status – re-check all dependency availability without restarting.
Open Folder – open the current runtime’s plugin-managed dependency directory.
Clear Current Runtime – remove all plugin-managed packages for the active QGIS runtime.
Clear All Runtimes – remove plugin-managed packages across all QGIS runtimes.

Warning

Clearing dependencies removes all plugin-installed packages. You will need to reinstall them before using Geo-SAM again.