To address the challenges of high labor intensity, low efficiency, and growing labor shortages in Morchella harvesting, this paper proposes the YOLO-CAS algorithm and develops a corresponding maturity detection system. Built upon the YOLOv8s architecture, the proposed model integrates the Convolutional Block Attention Module (CBAM) and Alterable Kernel Convolution (AKConv) while replacing the CIoU loss function with SIoU. These enhancements prioritize the refinement of feature extraction and selection for key Morchella characteristics, providing a robust algorithmic core for the detection system. The specific technical improvements are threefold: (1) CBAM modules are integrated into the three C2f-to-Detect coupling interfaces and the multi-scale output paths of the neck; (2) AKConv is embedded into the C2f structure at the end of the backbone prior to the SPPF module; and (3) the original CIoU is substituted with the SIoU loss function. Experimental results demonstrate that the refined model’s precision increased from 0.847 to 0.939, recall rose from 0.761 to 0.867, mAP50 improved from 0.853 to 0.927, and mAP50:95 advanced from 0.531 to 0.599, alongside a boost in inference speed. Furthermore, a deployment system with a streamlined and intuitive interface was developed using PyQt5. This system offers high flexibility and scalability, effectively meeting the practical demands of diverse agricultural environments.
Wang et al. (Wed,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: