WeeDB will provide a curated image database for AI-based weed detection in directly sown vegetable crops, with expert-supported annotations of weed plants and their apical meristem centres. The use case builds on the JaetRobi project, which aims to develop practical, herbicide- free, automated weed regulation in directly sown vegetable crops using sensor technology, image databases, AI-based weed recognition, and robotic actuation. Introduction Weed control in organic and direct-sown vegetable production is highly labour-intensive, especially in slow-developing crops such as carrot, beetroot, onion, and herbs, where weeds must be removed as early as possible around the crop plants. Methodologies The overarching aim is to provide data, annotation workflows, and benchmark-ready resources for AI-based weed detection and meristem-centre localisation. • Collect image data of crop–weed scenes from directly sown vegetable crops under practical field conditions, particularly focusing on early weed development stages. • Use a dedicated annotation tool and detection models developed to support combined bbox + meristem-centre annotation. • Link AI outputs to robotic or mechanical weed-control modules, following the JaetRobi goal of combining sensor and actuator components into functional weeding units. Output • Curated weed image database for directly sown vegetable crops with Apical-meristem-centre annotations • Annotation tool for combined bbox and meristem-centre labelling. • Standardised metadata describing crop type, growth stage, imaging setup, field conditions, annotation type, and quality-control status. • Baseline AI models for weed detection and meristem-centre localisation. What I can offer • An annotation tool for combined bounding-box and meristem-centre annotation. • Experience with AI-ready annotation workflows for plant detection tasks. • Curated crop–weed image data from practical field conditions. • Domain expertise in weed detection, early-stage plant recognition, and AI model validation. What I need • Data and partners working on AI models for object detection, keypoint localisation, and transfer learning. • Robotics and actuator partners interested in using meristem-centre outputs for targeted weed removal. • Collaboration opportunities in national or EU projects on herbicide-free automated weed regulation.

Source: Annual AgData Use Case Summit 2026 booklet.