Thank you for helping improve agricultural machine learning. AgML welcomes contributions from technical and agricultural-domain communities. You do not need to contribute code.
Start in Project-AgML Discussions. Include:
- The agricultural system, crop, region, or problem
- The people who would use the result
- Available data modalities and approximate scale
- What you have already tried
- What you can contribute
- The decision, standard, tool, or collaboration you need
Preliminary ideas are welcome. Please distinguish established results from work in progress.
Use Datasets, Metadata & Standards to discuss:
- Missing public datasets
- Dataset provenance, licensing, citations, and attribution
- Crop, geography, season, phenology, sensor, and platform metadata
- Label ontologies and regional or scientific naming
- Soil, weather, management, sensor, harvest, or other multimodal data
- Quality, duplicates, leakage, annotation reliability, and representativeness
Do not upload private, sensitive, restricted, or personally identifying data. A public data dictionary or de-identified example schema is often enough to begin.
Use Models, Benchmarks & VLMs to propose an evaluation or share preliminary findings. For results intended for the public leaderboard, follow the leaderboard contribution instructions.
Please report enough information to reproduce and interpret the result, including the model and version, dataset and configuration, task, split, prompt or training setup, metrics, hardware when relevant, and known limitations.
Scoped bugs and implementation tasks belong in Issues in the relevant repository:
- AgML Python library
- AgML website and leaderboard
- AgML Community for community documentation and operations
Before beginning a substantial implementation, open or join a discussion so maintainers and potential collaborators can help clarify scope.
Maintainers will help synthesize mature discussions into a clear problem statement, evidence, proposed next step, and appropriate repository. A discussion may then become an issue, documentation change, benchmark run, dataset update, research experiment, or working group.
Not every discussion will become an AgML feature. When a proposal is deferred or declined, maintainers should explain the relevant constraint or tradeoff.
Follow the Code of Conduct. Be specific, constructive, and generous with context. Credit data creators and prior work. Avoid presenting preliminary findings as validated conclusions.
AgML will credit substantive community contributions in the relevant discussion, issue, pull request, dataset or model card, release notes, and public communications when appropriate. Tell maintainers how you prefer to be credited.