KI-PlanetFunded by the Bavarian State Ministry for Science & Arts

Source: https://bit.cs.tum.de/en/research/projects/rl4flowsheet-1-1 Parent: https://bit.cs.tum.de/en/

Project Description

With climate change and the push to reduce chemical agents in agriculture, such as the Bavarian government's goal to cut pesticide use by 50% by 2028, breeding disease-resistant crops is increasingly vital. Modern plant breeding is a lengthy, complex, and expensive process that must account for environmental factors, pathogens, and fertilisation. Accurate monitoring of plant health is crucial for selecting resilient varieties that can withstand both biotic and abiotic stressors, such as fungal infections.

Genomic methods, particularly QTL mapping and GWAS, help breeders identify genome regions associated with desirable traits such as disease resistance. However, phenotyping, which consists of measuring plant traits, is a bottleneck in these studies. Current visual assessments, often done using a 1-9 scale, are prone to error and lack the precision needed for high-resolution genetic mapping. Additionally, certain diseases, such as Ramularia leaf spot, are difficult to visually differentiate, and manual phenotyping limits the scope and frequency of measurements.

Advances in machine learning, especially neural networks like CNNs, have revolutionised automated image analysis. Despite this, most existing AI models focus on disease classification rather than precise phenotypic measurement, which breeders need for disease severity assessment and QTL mapping. Moreover, further research is needed in the direction of AI-based segmentation models to accurately phenotype diseases such as Rhynchosporium leaf spot or Ramularia leaf spot.

This project aims to develop adaptable, open-source, AI-driven methods for image-based phenotyping of leaf diseases in barley. Focusing on diseases such as Brown rust and Ramularia leaf spot; the project will apply AI to accurately measure disease severity, essential for breeding resistant crops. In the third year, the methods will be validated for Rhynchosporium leaf spot.

The project will also create a browser-based dashboard for breeders to analyse their image data. By enhancing phenotyping accuracy through AI, this research will contribute to the accelerated development of disease-resistant crops, aligning with the global push for sustainable agriculture.

Project Information

Project title : Accurate AI-based phenotyping of climate stress-related leaf diseases for more efficient resistance breeding (KI-Planet)

Funding : The project is supported by funds of the Bavarian State Ministry of Science and Arts.