Orthoseg CLI Reference#

Orthoseg exposes a small set of command line tools for the main project workflow. All commands are installed as console scripts when the package is installed.

Common usage pattern#

Most commands use the same invocation pattern:

orthoseg_<command> --config path/to/project.ini [section.key=value ...]

The --config argument points to the project configuration file to use. Most commands also accept optional config overrules as positional arguments in the form <section>.<key>=<value>. This is useful for small one-off changes, such as predicting on another image layer without creating a separate .ini file.

Typical workflow#

The most common sequence is:

  1. orthoseg_load_sampleprojects to download the example projects (once).

  2. orthoseg_load_images to fetch or prepare imagery into the local cache (once per prediction layer).

  3. orthoseg_validate to prepare and validate training data (optional, validate also runs automatically at the start of orthoseg_train).

  4. orthoseg_train to train a model.

  5. orthoseg_predict to run inference on an image layer.

  6. orthoseg_postprocess to further postprocess the prediction output.

Command reference#

orthoseg_load_sampleprojects#

Downloads the sample projects repository into a local orthoseg/sample_projects directory.

Usage:

orthoseg_load_sampleprojects DEST_DIR [--ssl_verify true|false|PATH]

Important arguments:

  • DEST_DIR: base directory where the orthoseg/sample_projects directory will be created.

  • --ssl_verify: controls certificate validation for the download. It accepts true, false, or a path to a certificate bundle.

Use this command to get a working example project structure, sample data, and a pretrained model for the football fields example.

orthoseg_load_images#

Loads images for a layer into cache directories for orthoseg_predict.

The images will follow the tiling scheme configured in the image layer and project configuration. Normally the images will have a certain overlap to avoid edge effects in the predictions.

Usage:

orthoseg_load_images --config path/to/project.ini [section.key=value ...]

Important arguments:

  • --config: the project configuration file.

  • section.key=value: optional configuration overrules.

orthoseg_validate#

Validates the training data defined by the project configuration.

This is useful if you have a training or prediction running already and you don’t want to interrupt it to validate another training dataset.

Usage:

orthoseg_validate --config path/to/project.ini [section.key=value ...]

Important arguments:

  • --config: the project configuration file.

  • section.key=value: optional configuration overrules.

orthoseg_train#

Runs a training session for the configured project.

This command prepares output directories, reads the configured training labels, and trains the configured model.

Usage:

orthoseg_train --config path/to/project.ini [section.key=value ...]

Important arguments:

  • --config: the project configuration file.

  • section.key=value: optional configuration overrules.

orthoseg_predict#

Runs inference for the configured project on the configured predict.image_layer.

It uses the newest trained model and writes prediction outputs as vector features with some basic postprocessing being applied on the fly already.

Usage:

orthoseg_predict --config path/to/project.ini [section.key=value ...]

Important arguments:

  • --config: the project configuration file.

  • section.key=value: optional configuration overrules.

Example:

orthoseg_predict --config footballfields.ini predict.image_layer=BEFL-2020

orthoseg_postprocess#

Post-processes the prediction output.

Especially postprocessing steps that cannot be applied on the fly on individual prediction tiles are applied in this step. For example, if you want to merge adjacent polygons that are predicted in different tiles with a dissolve operation, this can be applied here.

Usage:

orthoseg_postprocess --config path/to/project.ini [section.key=value ...]

Important arguments:

  • --config: the project configuration file.

  • section.key=value: optional configuration overrules.

orthoseg_postprocess is normally run after orthoseg_predict.

osscriptrunner#

Runs job scripts from a directory and can optionally keep watching that directory for new work. This is the automation-oriented entry point for scheduled or unattended processing.

Usage:

osscriptrunner --script_dir path/to/jobs [--watch] [--config path/to/scriptrunner.ini]

Important arguments:

  • --script_dir: directory containing scripts or job files to execute.

  • --watch: keep monitoring the directory for new jobs.

  • --config: optional scriptrunner configuration file that overrides defaults.

The scriptrunner is separate from the project .ini files used by the other CLI commands. It is intended for orchestration rather than for a single train or predict run.