Healthy soils are fundamental to global food production, yet agricultural productivity is increasingly threatened by soil degradation, intensive pesticide use, and plant diseases, which are estimated to cause 20–40% of global crop losses. Improved methods for monitoring soil microbial communities and detecting plant pathogens before disease development are therefore urgently needed. This project develops an innovative framework combining transparent microfluidic soil chips, microscopy, and deep learning-based object detection to enable direct, image-based analysis of microorganisms in soil. We have previously demonstrated that this approach can be applied across laboratory incubations, in situ observations, and field experiments, enabling quantification of bacterial abundance and morphological variation across geographical gradients. By integrating high-resolution phenotypic information with field observations, the framework complements conventional molecular and culture-based methods for studying soil microbial diversity and dynamics. Building on these developments, we are now shifting our focus towards fungi and oomycetes, particularly soil-borne plant pathogens that have major impacts on agricultural productivity and crop yield. We are developing advanced deep learning models for the detection and quantitative analysis of fungal hyphae and pathogen-associated morphological features. Over the coming year, we aim to establish a database containing at least 15,000 images of plant-pathogenic soil-borne fungi and oomycetes for training and validating object detection models, with continued expansion in subsequent years. The adaptability of the framework to different microscopy platforms, including portable field microscopes, provides opportunities for large-scale and potentially global deployment. Ultimately, this approach aims to provide a foundation for early, image-based monitoring of soil-borne pathogens in agricultural fields. Earlier and more spatially resolved detection could support targeted disease management, reducing unnecessary pesticide application while helping prevent disease-associated crop losses. This work is part of the project Elucidating Spatiotemporal Dynamics of Soil Microbes via Belowground Microscopy, funded by the Swedish Research Council’s 2024 Interdisciplinary Research Environment Grant. NAISS computational resources will support the development and training of deep learning models for fungal and oomycete pathogen detection and large-scale image data analysis.