Tomato Maturity Detection and Quality Grading Dataset

(1) Everyone wants to purchase high-quality, fresh fruits. People nowadays are so concerned about their health and conscious about what they should consume and what they should not. According to them, rotten fruits are detrimental to their health. As a result, the sale of fruits suffers which brings significant economic ramifications. One of the main reasons behind fruits getting defective is that the maturity detection process is still manual in Bangladesh and fruits would go rotten with the passage of time if it is not harvested at the proper time. To understand the proper harvesting time it is crucial to detect the mature and immature fruit properly. Tomato is one of the significant, widely consumed, well-liked, and nutrient-dense crops grown throughout Bangladesh. According to the estimation, in Bangladesh, every day there are significant financial losses because tomatoes become rotten easily. Hence, automated classification of mature, immature, fresh, and rotten tomato identification is indispensable to overcome this situation and bring out the benefit to fruit growers, retailers, and processing firms.

(2) In the recent era, computer vision techniques are very promising in performing such types of classification and detection tasks.

(3) With a view to developing computer vision-based algorithms, an extensive tomato dataset is presented containing Tomato Maturity Detection Dataset and Tomato Quality Grading Dataset. Tomato Maturity Detection Dataset consists of two classes namely immature and mature tomatoes whereas Tomato Quality Grading Dataset contains fresh and rotten tomatoes. The classifications of this dataset are done with the help of a domain expert from an agricultural institute.

(4) A total of 2986 images of mature, immature, fresh, and rotten tomatoes were collected from Sher-e-Bangla Agricultural University. Then from these original images, a total of 10,000 augmented images are produced by using rotation, zoom, flipping, and scaling techniques to increase the data number.

(1) 每个人都希望购买高品质的新鲜水果。如今人们高度关注健康，对饮食选择有着清晰的认知——明确该吃什么、不该吃什么。在他们看来，腐烂的水果对健康有害，这导致水果销量受损，进而引发重大经济影响。孟加拉国水果变质的主要原因之一在于成熟度检测仍依赖人工，若未在适当时机采收，水果会随时间腐烂。因此，准确检测水果的成熟与未成熟状态，对于把握最佳采收时机至关重要。番茄是孟加拉国广泛种植的重要作物之一，因其营养丰富、消费量大且深受喜爱。据估计，孟加拉国每天因番茄易腐烂而遭受重大经济损失。因此，实现成熟、未成熟、新鲜及腐烂番茄的自动化分类，对于改善这一状况、为果农、零售商及加工企业带来效益而言，是不可或缺的。

(2) 近年来，计算机视觉技术（Computer Vision Techniques）在执行此类分类与检测任务方面展现出巨大潜力。

(3) 为开发基于计算机视觉的算法，本研究提出一个包含番茄成熟度检测数据集与番茄质量分级数据集的大规模番茄数据集。番茄成熟度检测数据集包含未成熟和成熟番茄两个类别，而番茄质量分级数据集则涵盖新鲜和腐烂番茄。该数据集的分类工作由来自农业研究所的领域专家（Domain Expert）协助完成。

(4) 数据集共从谢尔-伊-班格拉农业大学收集了2986张成熟、未成熟、新鲜及腐烂番茄的原始图像。随后，通过旋转、缩放、翻转及拉伸等技术对原始图像进行增强，生成了10000张增强图像，以扩充数据量。