GLiMPSe: A low-cost, high-throughput and accurate field phenotyping system for maize architectural traits

Maize phenotyping remains a major bottleneck in genetic analysis and breeding. Despite advances in drones, field robots, and gantry phenotyping systems, ultra-affordable, high-throughput, field-based maize phenotyping at single-plant resolution is still lacking, largely due to the high cost, complex deployment, and limited flexibility of existing platforms under heterogeneous field conditions. To address these challenges, we propose a novel paradigm that integrates DIY imaging devices with customized computer vision–based analytics, and present GLiMPSe ( G iraffe + Li zard M aize P henotyping S yst e m), including two end-to-end phenotyping modules for maize plant architecture (the Giraffe module) and leaf traits (the Lizard module). The imaging device in the Giraffe module are built from modular electronics and 3D-printed parts from local retailers to achieve high-quality image acquisition. The Giraffe and the Lizard modules operate at speeds of 15 seconds and 8 seconds per sample, with costs of $379.1 and $241.1, respectively. Both modules feature fine-tuned YOLOv11x segmentation models for reliable and robust target segmentation, followed by customized Python-based analytical pipelines that enable precise extraction and quantification of phenotypic traits. This methodology achieves high accuracies ( R² ) for five key traits, including plant height (0.928), heights of above-ear leaves (0.87∼0.958), ear height (0.925), above-ear leaf number (0.837), and leaf width (0.937). To enhance accessibility, we developed user-friendly graphical interfaces and publicly released manually annotated datasets and source code to support broader adoption and further innovation. This work provides a practical and accessible solution for high-throughput field phenotyping and offers new opportunities for democratizing crop phenomics through affordable, open-source technologies.