Vision for Robotfusion

NLAI takes part as a project participant and contributes engineering and data work:

• AI model validation — hands-on validation of computer vision models against shared, multi-source image data.
• Image transformation — pipelines for normalising, transforming and preparing image data for downstream model training.
• Dataset development — structured dataset preparation, annotation review and quality checks for reusable assets.
• Tooling support — engineering support for tooling that makes AI workflows reproducible and inspectable.
• Data-sharing methodology — practical methods for responsible data sharing across organisations and use cases.
• Community documentation — documentation that helps the broader Brabant AI community adopt and reuse what is built.

The work follows the path the data takes, and each step builds on the one before:

1. Prepare the data once. Images contributed by several organisations are transformed, normalised and structured into documented datasets, and their annotations are reviewed and quality-checked.
2. Test whether sharing helps. Vision models for robotic and agricultural use are developed and validated on the pooled, multi-source data, to establish whether sharing measurably improves them.
3. Make sharing workable. Shared data structures and governance principles set out how organisations can pool data in a way that is practical and trustworthy.
4. Package what works for reuse. Tooling, templates and documentation are brought together into a generic approach that other Brabant.ai initiatives can adopt.