LeafTip-RN: Generative AI-powered temporal interpolation for continuous phenotypic analysis of seedling establishment...

Area of Science:

• Agricultural Science
• Plant Science
• Computational Biology

Background:

• Seedling establishment is crucial for crop yield, but manual assessment in large trials is impractical.
• Dynamic traits like growth rate and uniformity require continuous, scalable monitoring.

Purpose of the Study:

• To develop and validate an automated pipeline (LeafTip-RN) for dynamic phenotyping of wheat seedling establishment.
• To leverage deep learning and generative AI for accurate and scalable data analysis.

Main Methods:

• Utilized ultralow-altitude drone phenotyping for flexible data collection.
• Developed a deep learning model for automated leaf-tip feature extraction.
• Integrated generative AI to interpolate data and create an extensive training library (353,019 leaf tips).

Main Results:

• Successfully quantified agronomically important traits like leaf tips and spatial uniformity for 51 wheat varieties.
• Derived growth curves and classified varieties into performance groups, revealing discrepancies with manual assessments.
• Developed an accessible GUI for visualizing and analyzing seedling development.

Conclusions:

• LeafTip-RN offers a scalable, GenAI-powered solution for evaluating wheat seedling establishment.
• The pipeline provides valuable tools for breeders to identify varieties with enhanced early growth and emergence.
• The methodology is extensible to other cereal crops.