A novel dataset of guava fruit for grading and classification

Abstract

Machine learning algorithms play a vital role in object detection and recognition. Currently, Machine learning techniques have achieved significant performance in various areas. However, there is still a need for research in the agriculture sector. The fruit harvesting process is carried out by unskilled labour without using modern scientific technologies; resultantly, the accuracy of harvesting is compromised. Moreover, immature fruits were harvested, which caused revenue losses and pretended sustainable growth. Therefore, the classification and grading of fruits are increasingly highlighted amongst the research communities. This article presents a novel dataset for local varieties such as Local Sindhi, Thadhrami and Riyali of guava fruit harvested in the Larkana region of Pakistan. The dataset is a primary instrument for developing an autonomous system using machine learning and deep learning methods. Hence, it has come up with an indigenous and state-of-the-art dataset. The dataset was developed using varieties as mentioned above. The dataset has been classified into three folders; each folder was further divided into three subfolders related to maturity level (i) Green, (ii) Mature Green, and (iii) Ripe. Images have been acquired in a controlled environment. The proposed dataset contains 2,309 total images in jpg format. This dataset will contribute to developing machine learning-based systems for the agricultural sector.

Keywords

• Guava fruit
• Grading
• Machine learning
• Deep Learning