Postdoc position – Self-Supervised Learning for Image-Based Phenotyping

Future food systems urgently need new crops that deliver sustainability, nutrition, and resilience. This includes crops that support the protein transition and thrive under changing climates while staying within planetary boundaries. However, breeding progress on key traits, such as yield stability and drought tolerance, remains too slow to meet accelerating agricultural pressures. These traits are governed by complex genetic architectures and strong environmental interactions, making rapid improvement particularly challenging.

Digital phenotyping offers a transformative opportunity to accelerate crop improvement. Drones, satellites, and high-throughput sensors now generate large-scale, multi-temporal image data at single-plant resolution, capturing subtle variation in growth, stress responses, and yield components that conventional approaches cannot detect. The key challenge is no longer data collection but transforming these vast and complex datasets into actionable insights for advanced breeding.

In this context, the PHENOM project aspires to develop next-generation AI methods for plant phenotyping. The project focuses on self-supervised foundation models that learn biologically meaningful representations from plant images. Such models generate embeddings that could capture genetic variation, environmental responses, and their interactions, forming a new digital layer for phenotyping that goes beyond traditional statistical approaches.

As a postdoctoral researcher, you will design and develop self-supervised learning approaches for large-scale, multi-temporal plant image datasets. You will work on extracting robust and interpretable representations of plant traits, enabling more precise phenotyping and supporting predictive breeding pipelines. 

You will work with quinoa as both a model and target crop—a resilient, protein-rich species suited to sustainable agriculture. The project is a collaboration between Radicle Crops and Wageningen University, combining extensive UAV datasets, real-world breeding pipelines, and world-leading expertise in agricultural AI. Together, the partners build an end-to-end pipeline from data curation and model development to validation in practical breeding applications.