HyperPRI: A Dataset of Hyperspectral Images for Underground Plant Root Study

Here we present Hyperspectral Plant Root Imagery (HyperPRI), the first available dataset of RGB and HSI data for in situ, non-destructive, underground plant root analysis using machine learning tools. HyperPRI contains images of plant roots grown in rhizoboxes for two annual crop species – peanut (Arachis hypogaea) and sweet corn (Zea mays). Drought conditions are simulated once, and the boxes are imaged and weighed on select days across two months. Along with the images, we provide hand-labeled semantic masks and imaging environment metadata. HyperPRI may be applied to semantic segmentation, plant phenotyping, and drought resilience studies. The proposed dataset may also have transferable insights for other datasets containing thin object features among highly textured backgrounds. Dataset Features Red-green-blue (RGB) and hyperspectral imaging (HSI) data Temporal data for rhizoboxes - plants are monitored from seedling till they are reproductively mature. Thin roots as narrow as 1-3 pixels Highly texture soil background High-resolution spectral data with high correlation between channels Computer Vision Tasks Compute root characteristics (length, diameter, angle, count, system architecture, hyperspectral) Determine root turnover Observe drought resiliency and response Compare multiple physical and hyperspectral plant traits across time Investigate texture analysis techniques Segment roots vs. soil