LeafScans-Orchard dataset

LeafScans-Orchard is an open RGB image dataset of orchard plant leaves designed for computer vision, machine learning, plant phenotyping, and agricultural image analysis. It contains 9708 flatbed scans of individual leaves acquired during five collection years between 2015 and 2025. It covers seven orchard crops, i.e. apple, pear, sweet cherry, sour cherry, plum, peach, and apricot, and includes 67 cultivar labels. The images were acquired under controlled conditions using flatbed scanning on a uniform background. Original scans were captured at 1200 dpi and released as lossless 300 dpi TIFF images. A representative subset of 300 original 1200 dpi scans is also provided for high-resolution analyses. The dataset is accompanied by image-level metadata. LeafScans-Orchard is suitable for species classification, cultivar recognition, fine-grained visual classification, leaf morphology and texture analysis, class imbalance studies, and evaluation of model robustness across acquisition years. It is released under the CC BY 4.0 license.