The project "Multimodal Lightweight and Explainable AI for Plant Health Diagnostics and Evidence-Linked Decision Support" (MLXAI-Plant) will develop a multimodal large language model capable of analysing plant images from Germany and Taiwan and linking its assessments to scientific literature. The system will identify plant species, detect diseases as well as stress factors, and provide evidence-based recommendations derived from peer-reviewed publications. A central innovation is the integration of explainable AI methods that generate transparent, verifiable, and legally compliant diagnostic reasoning. The project includes research on legal requirements for trustworthy AI, ensuring compliance with the EU AI Act, IP-law including trade secrets, data governance regulations, and cross-border data transfers. Additionally, lightweight Edge-AI variants will be developed for deployment on agricultural robots, protected cultivation platforms, and field systems. The project strengthens German-Taiwanese research collaboration and contributes to resilient, sustainable agricultural production systems.
Partners
- National Chung Hsing University (NCHU)
- University of Osnabrück (UOS)
- Deutsche Saatveredelung AG (DSV)
