YOLOV8-DFRE: A Lightweight Daylily Maturity Real-Time Detection Algorithm

Daylily is a versatile and economically valuable crop that is widely cultivated in China. Object detection of daylily is a key technology to realize intelligent harvesting in agriculture. In order to solve the problems of low efficiency and high subjectivity of traditional manual picking, a YOLOV8-DERE daylily maturity detection model based on YOLOv8n was proposed. In this method, the lightweight network model FasterRep module is combined with the C2f module to reduce the number of network parameters and improve the detection speed of daylily. The efficient multi-scale attention (EMA) moudle of cross-scale learning is introduced into the backbone network to enhance the network’s ability to extract multi-scale features and reduce the interference of complex backgrounds on the recognition effect. and improve the detection accuracy. Experimental results show that compared with the yolov8n model, the improved YOLOv8-DFRC model has an accuracy increase of 1.9%, a recall rate of 0.8%, a mAP@50 increase of 0.8%, an inference speed of 1.1ms, and a decrease of 22.6% and 21% in the number of parameters and model size, respectively. It meets the requirements of real-time detection and lightweight application, and provides a fast and accurate target recognition reference for the realization of automatic daylily picking.