Detection and Identification of Daylily Maturity Based on YOLOv8

To better apply object detection and identification techniques from deep learning to the field of agricultural automation, this paper focuses on the growth process of daylilies.It employs the state-of-the-art YOLOv8 model to achieve accurate assessment of daylily maturity.The backbone network of YOLOv8 draws inspiration from the CSPDarkNet network structure to extract image features.The introduction of cross-stage connections in the middle layers of the network enhances the efficiency of feature propagation.The neck section adopts the PAFPN bidirectional channel network to further process the features extracted by the backbone network, facilitating the smooth transmission of information from the bottom to the top layers.The Head section utilizes the Decoupled-Head structure to generate the output for object detection.The trained model had a mAP which reached up to 94%, an accuracy rate up to 95.6%, a recall rate up to 90.5%.Moreover, the identification speed is significantly improved.