Original dataset for "An Improved YOLOv8-based Lightweight Approach for Orange Maturity Detection

The dataset comprised two main components, totaling 1113 images and 5628 orange samples (He, 2024). The primary component consisted of 338 images depicting orange from the 'Chunjian' variety. These images were captured under varying conditions of light, distance, occlusion, number of targets, and maturity in Ziyan Township, Nanchong City, Sichuan Province, China. This variety is extensively cultivated in southern regions of China and represents one of the primary cultivars of citrus in the country. The image was captured using a Canon EOS 90D camera with an 18-55mm focal length lens, under natural outdoor lighting conditions. The camera was set to automatic exposure mode, and the processed image was exported with a resolution of 2992×2992 pixels.The secondary component of the dataset comprised 775 images sourced from publicly available orange picture on the internet, totaling 1216 samples. These additional images aim to enhance the model's generalization performance to other orange varieties.The images have been partitioned into training set (70%, 781 images used for model training), validation set (20%, 221 images used for hyperparameter tuning and performance evaluation during training), and test set (10%, 111 images used for assessing the model's generalization ability, accuracy, and reliability).