A faster and lighter weight robotic ready model YOLO Punica for detecting pomegranate fruit development

摘要

Pomegranate is a highly valued fruit known for its nutritional seeds packed with health benefits, as well as its use as a popular ornamental tree and in traditional medicine. However, traditional management relies heavily on manual, laborious processes, leading to low efficiency and increased cost. Using machine vision to monitor fruit development in real-time can achieve accurate intelligent management and reduce computational costs. A fast and lightweight detection model is crucial for machine vision. This study proposes the YOLO-Punica model, built on an improved version of the You Only Looking Once version 8n (YOLOv8n) algorithm, resulting in a lightweight and faster detection model specifically designed to monitor pomegranate fruit development in real time. The optimization includes the integration of two innovative modules: a dual-path downsampling module (DPDM) and a cross-scale feature fusion module (CCFM). The incorporation of DPDM into the backbone network significantly enhanced detection precision and computational efficiency. Additionally, the integration of CCFM and DPDM into the neck structure substantially reduced parameters, memory consumption, and overall model size, while improving operational efficiency and detection accuracy. The implementation of the DPDM and CCFM in the YOLOv8n framework resulted in a lighter model, faster processing speeds, and improved detection accuracy. Comparative test results on public datasets indicated that YOLO-Punica achieved reductions of 45.8% in parameters, 28% in giga floating-point operations per second (GFLOP), and 43.7% in model size relative to YOLOv8n, while attaining a mean average precision (mAP) of 92.6%, surpassing YOLOv8n by 0.98%. Furthermore, the model processes images at 14.3 frames per second on embedded devices, demonstrating its applicability for real-time detection of pomegranate fruit development, even in low-computational-power environments. This research not only provides technical support for the intelligent detection of pomegranate fruit development but also provides a new perspective for enhancing machine vision models in other agricultural contexts. Our code and model are available at https://github.com/Wenxuan-889/yolo_Pomegranate .