Morphological Dataset of Indian Mango Leaf Varieties (MIT-WPU Campus)

Mango Leaf Morphological Dataset (Version 1.0) 1. Project Overview and Site Context This dataset comprises a systematic morphological study of 16 different mango (Mangifera indica) cultivars. All samples were collected from the MIT World Peace University (MIT-WPU) Campus, Kothrud, Pune, Maharashtra, India ($18.5177^\circ\text{ N}, 73.8148^\circ\text{ E}$). The site is situated at an altitude of 560m in a tropical wet and dry climate zone, with soil composed primarily of weathered Deccan Trap basalt. 2. Sampling and Acquisition Sample Size: 160 independent leaf samples (10 leaves per tree across 16 trees). Selection Criteria: Mature, healthy leaves were selected from the middle of the tree canopy to ensure representative varietal characteristics. Standardisation: Leaves were photographed against a flat, neutral background to minimise parallax error and distortion. 3. Highlight: Biological Ground Reference and Calibration To achieve high metric accuracy without specialised laboratory equipment, this dataset utilises a Human Hand Ground Reference system: Calibration Protocol: Every image contains a human hand placed in the same focal plane as the leaf sample. Metric Standard: The average adult male palm width measured at 82–85 mm serves as the fixed calibration constant. Accuracy: By calculating the pixel-to-millimetre ratio based on the reference hand, the Length and Breadth measurements provided in the metadata are physically accurate representations of the leaf’s true scale. This methodology allows for the training of Computer Vision models for Real-World Size Estimation. 4. Morphological Parameters Recorded The accompanying CSV metadata includes the following botanically significant data points: Leaf Length (mm): From the petiole base to the tip of the apex. Leaf Breadth (mm): The maximum width of the leaf blade perpendicular to the midrib. Variety Identification: Cross-referenced with local agricultural experts and IBPGR descriptors. Visual Markers: Capture of specific traits including marginal undulation (waviness), apex shape (acute vs. acuminate), and venation angles. 5. Technical Validation for Researchers The dataset is intended for use in: Plant Phenotyping: Using morphological data to differentiate between similar cultivars (such as Alphonso vs. Kesar). Deep Learning: Training Convolutional Neural Networks (CNNs) for automated variety classification. Agricultural AI: Developing mobile-based tools for farmers to identify varieties in the field using relative scale (hand-reference).芒果叶片形态数据集（版本1.0） 一、项目概况与采集场地背景 本数据集系统开展了16个不同芒果（Mangifera indica）品种的叶片形态学研究。所有样本均采集自印度马哈拉施特拉邦浦那市科思鲁德的MIT世界和平大学（MIT-WPU）校园，地理坐标为$18.5177^circ ext{ N}, 73.8148^circ ext{ E}$。该场地海拔560米，属于热带干湿气候区，土壤主要由风化的德干暗色岩（Deccan Trap basalt）构成。 二、样本采集与获取 样本规模：共160份独立叶片样本，16棵果树每棵采集10片叶片。 筛选标准：从树冠中部选取成熟健康的叶片，以确保品种特征具有代表性。 标准化处理：所有叶片均在平坦中性背景下拍摄，以最大限度减少视差误差与图像畸变。 三、核心亮点：生物基准参考与校准方法 为在无需专用实验室设备的前提下实现高精度度量，本数据集采用人体手部基准参考系统： 校准流程：每张图像中均放置一只与叶片样本处于同一焦平面的人手。 度量基准：以成年男性手掌平均宽度82–85 mm作为固定校准常数。 精度保障：通过参考人手计算像素与毫米的转换比例，元数据中记录的叶片长度与宽度数值均为叶片真实尺寸的精准表征。该方法可用于训练可实现真实世界尺寸估算的计算机视觉（Computer Vision）模型。 四、记录的形态学参数 配套的CSV格式元数据包含以下具有植物学意义的数据项： 叶片长度（mm）：从叶柄基部至叶尖的距离。 叶片幅宽（mm）：叶片垂直于中脉的最大宽度。 品种鉴定：经当地农业专家结合IBPGR品种描述标准交叉核验确认。 视觉标记特征：记录了多项特定性状，包括叶缘起伏（波浪状）、叶尖形态（急尖与渐尖）以及叶脉夹角。 五、面向研究人员的技术验证说明 本数据集适用于以下研究方向： 植物表型组学（Plant Phenotyping）：利用形态学数据区分相似品种（如阿方索芒果与凯萨尔芒果）。 深度学习（Deep Learning）：训练卷积神经网络（Convolutional Neural Networks, CNNs）实现品种自动分类。 农业人工智能（Agricultural AI）：开发面向农户的移动应用工具，借助相对尺度参考（人手参考）在田间实现品种识别。