'BSQG-1372: A Benchmark Dataset for Brasenia schreberi Quality Gradin

BSQG-1372 is a benchmark image dataset for automated quality grading of Brasenia schreberi (water shield) buds, supporting non-destructive, real-time post-harvest sorting research. It contains 1,372 RGB images (1,098 train / 138 val / 136 test) collected under realistic sorting-line conditions, with 23,550 annotated instances across two quality classes: Good (15,227) and Bad (8,323). Annotations were produced by three trained experts (Fleiss' κ = 0.86). Used to train and evaluate BSQG-YOLOv6n, a lightweight real-time detector for embedded edge deployment. Data collection and annotation: Nayanajith, J. Supervision: Zhao Zhang. See the associated publication for the full author list and methodology. License: CC-BY-4.0