A dataset of hairy peach fruit images at different ripeness levels for machine vision in smart orchards

With the development of smart agriculture, precise fruit detection and ripeness classification have become one of the key technologies in intelligent orchard management. In this paper, a dataset of high-quality ripeness classification images for hairy peach to enhance the application effectiveness of object detection models for machine vision in smart orchards. The dataset is based on real orchard scenes in China, using smartphones to capture images of hairy peach fruits at different ripeness levels, covering three categories: immature, semi-mature, and mature. To ensure data diversity, various factors such as different shooting distances, lighting conditions, and fruit growth states were considered during the collection process. After excluding images that are too dark, blurry, defocused, heavily occluded, and highly repetitive, high-quality and representative samples are manually selected according to fruit clarity, composition integrity, and balanced ripeness. A total of 1,245 high-resolution images (original dataset 8.23 GB, augmented dataset 15.56 GB) were obtained, and 3,493 fruit bounding boxes were manually annotated. The dataset provides annotation files in both Pascal VOC and YOLO formats. This dataset performs well with YOLO v8, RT-DETR, and Mamba-YOLO, and provides fundamental data support for the development of visual intelligent system for robots in smart orchard.