agri-image-meta 1.0.0

set of functions for common deeplearning tasks.

Agriculture Image Metadata

A Linked Data–based metadata model for agricultural image data powered by Vision+Robotics

This repository provides an ontology and python package for image dataset in agricultural domains (greenhouse, open field, arable crops, horticulture, phenotyping).

The goal is to offer a lightweight, interoperable, FAIR‑friendly schema that leverages existing ontologies and codelists, while adding only minimal project‑specific extensions where necessary.

Canonical namespace & hosting

• Namespace: https://w3id.org/agri-image/
• Ontology: ontology/ontology.ttl will be hosted at https://w3id.org/agri-image

Repository structure

## examples
├── examples
│   ├── custom_dataset_example.py
│   ├── filtering_class_based.py
│   ├── filtering_query_based.py
│   ├── your_custom_dataset.json
│   └── your_custom_dataset.yaml
├── metadata_vision ## Package for loading etc
...
├── ontology
│   ├── dummy_dataset_output.json
│   ├── dummy_dataset_output.ttl
│   └── ontology.ttl
LICENSE

Metadata Structure

The idea is to create a dataset that can consist of:

• fields: consisting of
  • plots/rows
    • which have crops and weeds
    • properties of:
      • soilType
      • surfaceLayer
      • weatherConditions Data is collected by a machine/platform
• platform: has sensors:
  • Camera: unique cam_id and properties
  • Lidar: To be implemented in near future
  • GPS: To be implemented in near future
• images: list of images which links to field, plot, platform and sensors

# DatasetMetadata(RDFModel)
#     hasField: list[FieldMetadata] | FieldMetadata
    # hasPlot: PlotMetadata
    # plot.hasCrop: CropMetadata | list[CropMetadata]
    # plot.hasWeed: CropMetadata | list[CropMetadata]
    # platform: PlatformMetadata list[PlatformMetadata]
    # platform.hasSensor: list[CameraMetadata] | CameraMetadata]
    # images: ImageMetadata | list[ImageMetadata]

Install for deployment

pip install metadata_vision@git+https://github.com/TeamWalabi/agriculture-image-metadata.git

Folder structure for deployment

We recommend following folder structure:

dataset_name / raw_data / field_id / plot_id / platform_id / cam_id / optional[YYYYMMDD]
# although for some projects it make sense to use:
dataset_name / raw_data / field_id / plot_id / platform_id / YYYYMMDD / cam_id
## inside the folder cam_id you have
## XXX.png files and XXX.metadata.json files describing ImageMetadata
# to facilitate with the folder structure have a look at utils.file_system.py

This looks complicated but this make is ideal for:

• Timeseries
• Drone data with field and plots
• Multiple machines on a field.
• Machine is flexibile can be a:
  • harvesting robot
  • data platform
  • or in a greenhouse an device which has multiple camera's sensors

Recreate ontology

To recreate the ontology:

python3 metadata_vision/create_ontology.py

Commit / validating:

Install aditional packages

pip install -e .[test]
## validate if everything is working correctly with
pytest tests/

License

This project is licensed under the Creative Commons Attribution 4.0 International (CC BY 4.0) license. See LICENSE.

You may use, modify, and distribute the contents as long as you provide attribution.

Contributing

Suggestions and pull requests are welcome, especially for improving ontology mappings or adding reusable examples.

Contact

• Bart van Marrewijk & Joep Tummers
• Wageningen Research
• Email: [e-mail]