BDMediHerb: Medicinal Herbal Fruits & Seeds Dataset of Bangladesh

𝑶𝒗𝒆𝒓𝒗𝒊𝒆𝒘:
Herbal fruits and seeds are naturally abundant and provide a wide range of medicinal and therapeutic benefits. They are widely used as remedies for many common and chronic diseases. However, manual identification requires expert knowledge and can be labor-intensive.
To address this challenge, the Medicinal Herbal Fruits & Seeds Dataset was developed to enable automatic identification using Machine Learning and Deep Learning techniques. This dataset aims to make herbal knowledge more accessible by supporting the development of intelligent classification systems.

𝑪𝒐𝒏𝒕𝒆𝒙𝒕 & 𝑴𝒐𝒕𝒊𝒗𝒂𝒕𝒊𝒐𝒏:
* This dataset comprises high-resolution images of herbal fruits and seeds. It is structured into two parts: the Original Dataset and the Augmented Dataset.
* This dataset holds nineteen of these common yet very useful herbal medicinal fruits and seeds for people to classify easily.
* The nineteen fruits and seeds included are: 𝗕𝗲𝗹𝗹𝗲𝗿𝗶𝗰 𝗠𝘆𝗿𝗼𝗯𝗮𝗹𝗮𝗻 (Bohera), 𝗕𝗹𝗮𝗰𝗸 𝗖𝘂𝗺𝗶𝗻 (Kalo jira), 𝗕𝗹𝗮𝗰𝗸 𝗣𝗲𝗽𝗽𝗲𝗿 (Gol morich), 𝗖𝗮𝗿𝗱𝗮𝗺𝗼𝗺 (Elachi), 𝗖𝗵𝗲𝗯𝘂𝗹𝗶𝗰 𝗠𝘆𝗿𝗼𝗯𝗮𝗹𝗮𝗻 (Horitoki), 𝗖𝗶𝗻𝗻𝗮𝗺𝗼𝗻 (Daruchini), 𝗖𝗹𝗼𝘃𝗲 (Lobongo), 𝗖𝘂𝗺𝗶𝗻 𝗦𝗲𝗲𝗱𝘀 (Jira), 𝗙𝗲𝗻𝘂𝗴𝗿𝗲𝗲𝗸 𝗦𝗲𝗲𝗱𝘀 (Methi), 𝗙𝗹𝗮𝘅 𝗦𝗲𝗲𝗱𝘀 (Tishi), 𝗚𝗮𝗿𝗹𝗶𝗰 (Rosun), 𝗚𝗼𝗼𝘀𝗲𝗯𝗲𝗿𝗿𝘆 (Amloki), 𝗚𝗶𝗻𝗴𝗲𝗿 (Ada), 𝗠𝗮𝗰𝗲 (Joyetri), 𝗡𝘂𝘁𝗺𝗲𝗴 (Joyfol), 𝗣𝘀𝗼𝗿𝗮𝗹𝗲𝗮 𝗙𝗿𝘂𝗶𝘁 (Bokuchi fol), 𝗦𝘁𝗮𝗿 𝗔𝗻𝗶𝘀𝗲 (Tara mashla), 𝗦𝗲𝘀𝗮𝗺𝗲 𝗦𝗲𝗲𝗱𝘀 (Til), 𝗧𝘂𝗿𝗺𝗲𝗿𝗶𝗰 (Holud).
*The data is mainly for herbal fruits and seeds identification purposes. The data are primarily collected from eight different places, including the National Botanical Garden, Shonir Akhra Kacha Bazar, and Jahangirnagar University.

𝑫𝒂𝒕𝒂𝒔𝒆𝒕 𝑪𝒐𝒎𝒑𝒐𝒔𝒊𝒕𝒊𝒐𝒏:
~ Total Images: 19,000
• 3,800 Original high-resolution images
• 15,200 Augmented images
~ Captured in the vibrant, real-world ambiance with optimized lighting setup.
~ Images acquired using smartphones.

𝑯𝒆𝒓𝒃𝒂𝒍 𝑭𝒓𝒖𝒊𝒕 & 𝑺𝒆𝒆𝒅 𝑪𝒍𝒂𝒔𝒔𝒆𝒔:
~ 19 distinct classes covering a wide variety of herbal fruits and seeds.
~ Each class contains 200 original images and 800 augmented images.
~ Includes common and rare herbal fruits and seeds across different lighting and background conditions.

𝑺𝒖𝒎𝒎𝒖𝒓𝒚 & 𝑭𝒐𝒓𝒎𝒂𝒕:
* This dataset is well-suited for research in plant pathology, machine learning, and image classification tasks related to agriculture, classification, and crop health monitoring.
* Folder structure is provided in the README file.

概述：
草本果实与种子资源天然丰富，具备多样的药用与治疗功效，广泛用作多种常见及慢性疾病的治疗方剂。然而，人工鉴定需依托专业知识，且耗费人力。为解决这一难题，研究人员开发了药用草本果实与种子数据集（Medicinal Herbal Fruits & Seeds Dataset），以支持借助机器学习（Machine Learning）与深度学习（Deep Learning）技术实现自动鉴定。本数据集旨在通过助力智能分类系统的开发，使草本相关知识更易于获取。

背景与研究动机：
* 本数据集包含草本果实与种子的高分辨率图像，分为原始数据集与增强数据集两部分。
* 本数据集涵盖19种常见且极具实用价值的药用草本果实与种子，便于使用者开展分类任务。
* 所包含的19种果实与种子分别为：毗黎勒（Belleric Myrobalan，Bohera）、黑种草子（Black Cumin，Kalo jira）、黑胡椒（Black Pepper，Gol morich）、小豆蔻（Cardamom，Elachi）、诃子（Chebularic Myrobalan，Horitoki）、肉桂（Cinnamon，Daruchini）、丁香（Clove，Lobongo）、孜然籽（Cumin Seeds，Jira）、胡芦巴籽（Fenugreek Seeds，Methi）、亚麻籽（Flax Seeds，Tishi）、大蒜（Garlic，Rosun）、余甘子（Gooseberry，Amloki）、生姜（Ginger，Ada）、肉豆蔻衣（Mace，Joyetri）、肉豆蔻（Nutmeg，Joyfol）、补骨脂果（Psoralea Fruit，Bokuchi fol）、八角茴香（Star Anise，Tara mashla）、芝麻籽（Sesame Seeds，Til）、姜黄（Turmeric，Holud）。
* 本数据集主要用于草本果实与种子的鉴定任务，数据主要采集自8处场所，包括国家植物园、Shonir Akhra Kacha Bazar以及贾汉吉尔纳加尔大学。

数据集构成：
~ 总图像量：19000张
• 3800张原始高分辨率图像
• 15200张增强图像
~ 图像采集于充满生机的真实场景，采用优化的布光方案。
~ 图像均通过智能手机采集。

草本果实与种子类别：
~ 共19个独立类别，涵盖多样的草本果实与种子品类。
~ 每个类别包含200张原始图像与800张增强图像。
~ 涵盖不同光照与背景条件下的常见及珍稀草本果实与种子样本。

总结与数据格式：
* 本数据集适用于植物病理学、机器学习以及与农业、分类及作物健康监测相关的图像分类任务研究。
* 文件夹结构详见"README"文件。