Ferdaus Md Hasanul, Prito Rizvee Hassan, Ahmed Masud, Islam Mohammad Manzurul, Ali Md Sawkat, Ibrahim Muhammad, Rasel Ahmed Abdal Shafi, Islam Maheen, Jabid Taskeed, Rahman Md Atiqur, Rahoman Md Mizanur
Department of Computer Science and Engineering, East West University, Dhaka, Bangladesh.
Department of Agricultural Extension, Ministry of Agriculture, Bogura, Bangladesh.
Data Brief. 2025 Jul 21;62:111908. doi: 10.1016/j.dib.2025.111908. eCollection 2025 Oct.
The mango image dataset presented in this article contains clear and detailed images of the fifteen most common and popular mango () varieties in Bangladesh: Amrapali, Ashshina Classic, Ashshina Zhinuk, Banana Mango, Bari-4, Bari-11, Fazli Classic, Fazli Shurmai, Gourmoti, Harivanga, Himsagor, Katimon, Langra, Rupali, and Shada. The mango specimens were sourced from various fruit markets across six districts of Bangladesh, namely Rajshahi, Chapai Nawabganj, Satkhira, Panchagarh, Rangpur, and Dhaka, which are famous for popular mango cultivation and availability to ensure a wide geographic representation. To maintain the quality and uniformity of images across the dataset, the images were captured using a high-definition smartphone camera under a standardized and controlled environment. Overall, the full dataset contains a total of 28,515 images, where 5703 images are original (raw) and 5703 images are processed with a blend of both real and virtual backgrounds. The processed images were further augmented resulting in a total of 17,109 augmented images. This is done to enhance their utility for training machine learning and deep learning models, particularly for performing computer vision tasks such as object detection, classification, and segmentation. This augmentation includes transformations such as flipping, rotation, shearing, blurring, variation of brightness and exposure, and introduction of noise to simulate diverse real-world scenarios and improve model robustness. This dataset holds strong reuse potential across computer vision, agriculture, food processing, and biodiversity research. It supports automated mango variety identification, sorting, grading, and quality assessment in precision agriculture. It can also aid in breeding climate-resilient, high-yield mango varieties, enhancing food security and sustainable farming. Additionally, it facilitates studies on phenotypic diversity, genetic correlations, and regional trait comparisons. The dataset can help ensure traceability, authenticity, and quality assurance, improving supply chains and export potential. From a biodiversity standpoint, it contributes to documenting and conserving unique mango varieties.
本文介绍的芒果图像数据集包含孟加拉国十五种最常见且受欢迎的芒果（品种）的清晰详细图像：阿姆拉普利、阿什希娜经典、阿什希娜日努克、香蕉芒果、巴里 - 4、巴里 - 11、法兹利经典、法兹利舒尔迈、古尔莫蒂、哈里万加、希姆萨戈尔、卡蒂蒙、朗格拉、鲁帕利和沙达。这些芒果样本来自孟加拉国六个地区的各个水果市场，即拉杰沙希、查派纳瓦布甘杰、萨特希拉、潘查加尔、朗布尔和达卡，这些地区以种植和供应受欢迎的芒果而闻名，以确保广泛的地理代表性。为了保持数据集中图像的质量和一致性，图像是在标准化和受控环境下使用高清智能手机相机拍摄的。总体而言，完整数据集共有28515张图像，其中5703张是原始（未处理）图像，5703张是经过真实背景和虚拟背景混合处理的图像。经过处理的图像进一步增强，最终得到17109张增强图像。这样做是为了提高它们在训练机器学习和深度学习模型方面的效用，特别是用于执行诸如目标检测、分类和分割等计算机视觉任务。这种增强包括翻转、旋转、剪切、模糊、亮度和曝光变化以及引入噪声等变换，以模拟各种现实世界场景并提高模型的鲁棒性。该数据集在计算机视觉、农业、食品加工和生物多样性研究等领域具有很强的数据重用潜力。它支持精准农业中的芒果品种自动识别、分选、分级和质量评估。它还可以帮助培育适应气候变化、高产的芒果品种，增强粮食安全和可持续农业。此外，它有助于进行表型多样性、遗传相关性和区域性状比较的研究。该数据集有助于确保可追溯性、真实性和质量保证，改善供应链和出口潜力。从生物多样性的角度来看，它有助于记录和保护独特的芒果品种。
