A comprehensive tree leaf image dataset for morphometric studies

Abstract Despite similar universal primary physiological functions, plant leaves exhibit myriad shapes and sizes. Understanding this morphological variation is invaluable in plant taxonomy, ecology, evolution, and biomimetics. Achieving a comprehensive understanding of eco-evo-devo research requires diverse leaf-image datasets collected across regions and over time. While many datasets support morphometric studies using advanced imaging and machine learning, few provide standardised leaf images that enable uniform interspecific comparisons. We present a dataset of 161 high-quality RGB images of leaves of wild and cultivated tree species from Kerala, India, collected between 2020 and 2023. All leaves, including their petioles, were scanned using a digital scanner (Epson L360), centrally framed on a white background, and uniformly scaled to 1024 × 1024 pixels. In addition, the dataset comprises codes to compute the leaf morphometry using two novel objective morphometric measures: Segmental fractal complexity ( D ΣS ) and Geometric entropy ( S L ). These metrics were validated against the leaf dataset, showing strong correlations between D ΣS and leaf dissection index ( LDI ) ( ρ = 0.94) and between S L and D ΣS ( ρ = 0.94), confirming the relationship between leaf patterns and leaf lobiness, pinnation, and serration. D ΣS surpasses LDI by incorporating spatial positioning of leaflets, lobes and fine serration features. Both D ΣS and S L outperform geometric morphometric techniques, which are limited to intraspecific comparisons. Their objectivity, ease of use, and lack of statistical preprocessing make D ΣS and S L reliable metrics for interspecific leaf comparisons. We encourage researchers to expand or replicate our analysis using codes and leaf datasets from diverse locations. This dataset supports the development and validation of future leaf morphometric techniques. Despite limitations in high-resolution imaging and intraspecific variability, it remains valuable for advancing research and fostering collaboration across taxonomy, ecology, and computer vision.