ICPTC : Iranian Commercial Pistachios Tree Cultivars

This dataset includes 526 high-resolution images of 4 cultivars (Jumbo, Long, Round, and Super long) of Pistachio trees from their natural habitats and is called ICPCTC (Iranian Commercial Pistachios Tree Cultivars). From each cultivar, different numbers of images (109-171) of the entire trees and branches were captured in the natural habitat of the tree. In summary, out of 526 ICPTC images, 107 images were dropped for testing. About 80% of the images (419 remaining images) were used for training. In each folding, 86 images (about 20%) were used for validation and 333 images (about 80%) were used for training. The number of images from each class used for testing, validation, and training, was 20%, 17%, and 63% respectively. Images were collected from multiple Pistachio trees, at different camera-to-target distances and from different angles and viewpoints. The Pistachio is an important nut around the world. The pistachio tree has various cultivars that are financially important. These species play a significant role in the economy of Iran. The Pistachio tree cultivar recognition is very critical for cost-effective survival and sustainability. The Pistachio cultivar recognition in their natural habitats is a process. Automatic plant recognition by using image processing and computer vision can reduce the cost and time effectively. Modern computer vision approaches are based on deep learning techniques. In order to perform a deep learning algorithm, the existence of a standard dataset is very essential.本数据集命名为ICPCTC（伊朗商业开心树品种，Iranian Commercial Pistachios Tree Cultivars），包含采自自然生境的4个开心树品种（Jumbo、Long、Round、Super long）的526张高分辨率图像。针对每个品种，研究人员在其原生自然生境中采集了整株树木及枝条的图像，单品种图像数量在109至171张不等。 总体而言，526张ICPCTC图像中，107张被预留为测试集。剩余的419张图像（占总图像量的约80%）被用于模型训练。在每一轮交叉验证折叠中，86张图像（约占20%）被用作验证集，333张图像（约占80%）被用作训练集。此外，每个品种类别的图像在测试集、验证集和训练集中的占比分别为20%、17%和63%。所有图像均采自多株开心树，拍摄时相机与目标的距离各不相同，且涵盖了不同的拍摄角度与视角。 开心果是全球重要的坚果品类，开心树拥有多个具有重要经济价值的品种，这些品种对伊朗的国民经济发挥着关键作用。开心树品种识别对于实现低成本种植与产业可持续发展至关重要。在自然生境中开展开心树品种识别是一项具有实际意义的任务。借助图像处理（image processing）与计算机视觉（Computer Vision）技术实现植物自动识别，可有效降低成本并缩短耗时。当前主流的计算机视觉方法均基于深度学习技术，而要部署深度学习算法，标准数据集的构建是极为必要的基础条件。