New COMPAG Publication Introduces XylemVision for Data-Efficient Root Anatomy Analysis

We are excited to announce that our paper, "XylemVision: A Knowledge-Distilled Deep Learning Framework for High-Throughput Quantification of Root Anatomy in Peanut," has been accepted for publication in Computers and Electronics in Agriculture (COMPAG).

In this work, we developed XylemVision, a novel deep learning framework that combines knowledge distillation, object detection, and SAM-based segmentation to automate root anatomy analysis from microscope images while requiring only a small amount of labeled training data. The framework achieved high detection accuracy, produced precise segmentation masks for anatomical trait extraction, and demonstrated strong transferability across crop species, including soybean.

This research provides a scalable and data-efficient solution for high-throughput plant phenotyping and has the potential to accelerate root biology and crop improvement studies across diverse agricultural systems.