How worldcereal/worldcereal-classification works
WorldCereal is a Python system for producing cropland and crop-type maps from satellite and auxiliary data, primarily through cloud openEO jobs. A user specifies an area, time range, product, and optional season or model; the job manager prepares and runs work, and retrieves outputs. The repository also supports notebook-based reference-data extraction and custom model training, alongside local inference and STAC utilities. Edges are limited to relationships documented in the README or visible in the provided source index; several feature branches are shown without guessed internal wiring.
Main entry points: jobmanager.py Batch job manager · job.py Inference job · classification_app.py Classification app · extractions.py Extraction tools
Components 25
Job orchestration
- Job parameters
src/worldcereal/job_params.py - Task types
src/worldcereal/parameters.py - Batch job manager
src/worldcereal/jobmanager.py - Inference job
src/worldcereal/job.py
Mapping workflows
- Workflow presets
src/worldcereal/openeo/workflow_config.py - Satellite preprocessing
src/worldcereal/openeo/preprocessing.py - Seasonal inference
src/worldcereal/openeo/inference.py - CatBoost inference
src/worldcereal/openeo/inference_catboost.py - Local inference
src/worldcereal/local_inference.py - Season definitions
src/worldcereal/seasons.py
Training and data
- Training data
src/worldcereal/train/data.py - Model datasets
src/worldcereal/train/datasets.py - Seasonal training
src/worldcereal/train/seasonal.py
Notebook tools
- Classification app
notebooks/notebook_utils/classification_app.py - Notebook job tools
notebooks/notebook_utils/job_manager.py - Extraction tools
notebooks/notebook_utils/extractions.py - CDSE authentication
notebooks/notebook_utils/auth_utils.py
Reference data and catalogues
- Reference data access
src/worldcereal/utils/refdata.py - Reference Data Module
src/worldcereal/rdm_api/rdm_interaction.py - STAC interaction
src/worldcereal/stac/stac_api_interaction.py - Class legends
src/worldcereal/utils/legend.py - STAC catalogue builder
scripts/stac/catalogue_builder.py
Other components
- Map user
- openEO backend
- Downloaded results
Connections 22
- Map user → Job parameters: specifies
- Job parameters → Batch job manager: configures
src/worldcereal/jobmanager.py - Task types → Batch job manager: selects task
src/worldcereal/jobmanager.py - Batch job manager → Inference job: dispatches
src/worldcereal/jobmanager.py - Inference job → Workflow presets: configures
src/worldcereal/job.py - Inference job → Satellite preprocessing: uses
src/worldcereal/job.py - Inference job → Seasonal inference: invokes
src/worldcereal/job.py - Batch job manager → openEO backend: submits jobs
- Batch job manager → Downloaded results: downloads assets
src/worldcereal/jobmanager.py - Local inference → Seasonal inference: uses
src/worldcereal/local_inference.py - Classification app → Extraction tools: retrieves data
notebooks/notebook_utils/classification_app.py - Classification app → Notebook job tools: launches jobs
notebooks/notebook_utils/classification_app.py - Classification app → CDSE authentication: uses login
notebooks/notebook_utils/classification_app.py - Extraction tools → Reference data access: queries
notebooks/notebook_utils/extractions.py - Extraction tools → Reference Data Module: uses API
notebooks/notebook_utils/extractions.py - Training data → Model datasets: uses
src/worldcereal/train/data.py - Model datasets → Seasonal training: uses season logic
src/worldcereal/train/datasets.py - Reference data access → Class legends: uses labels
src/worldcereal/utils/refdata.py - STAC catalogue builder → STAC interaction: builds catalogue
- Map user → Training data: trains custom models
- Batch job manager → CatBoost inference: optional inference
- Batch job manager → Season definitions: resolves seasons
src/worldcereal/jobmanager.py