SUPR
Deep learning for soil chip
Dnr:

NAISS 2025/22-250

Type:

NAISS Small Compute

Principal Investigator:

Hanbang Zou

Affiliation:

Lunds universitet

Start Date:

2025-03-01

End Date:

2025-09-01

Primary Classification:

10606: Microbiology (Medical aspects at 30109 and agricultural at 40302)

Webpage:

Allocation

Abstract

Healthy soils lay the foundation for the world’s food production, but today soil health is at risk due to unsustainable agriculture practices such as the use of toxic pesticides. At the same time, we still lose 20-40% of our total yields due to plant diseases. Agriculture needs new methods to protect their crops from diseases. In this project, we develop new systems for early detection of the cause behind many plant diseases, plant pathogenic soil fungi and fungal like species, to stop them before they cause harm to the crops. Through the use of transparent microfluidic chips and deep learning driven image analysis, we enable visualizing microbe communities directly in soil. With this we can scan agricultural fields to find harmful microbes, and later guide farmers to more precise pest management, instead of blindly spraying the entire field with pesticides. This will enable farmers to decrease the use of pesticides, while simultaneously increasing crop yields by stopping diseases before they cause crop loss. We aim to build a database of at least 15’000 images of plant pathogenic soil-borne fungal and fungal like species over the coming year that will be trained on an object detection model. Over the few coming years, we aim to further develop and expand this database.