基于改进YOLOv8的番茄叶片病害智能检测算法OA

To address the challenges of early identification of tomato leaf diseases and the large parameter size of conventional YOLOv8 models that hinder edge deployment,an intelligent tomato leaf disease detection method integrating Ghost lightweight convolution and the CBAM attention mechanism is proposed.Based on the YOLOv8n framework,all C2f modules in the backbone are replaced with C3Ghost,and downsampling convolutions are replaced with GhostConv to reduce computational cost while maintaining feature extraction capability.CBAM modules are further inserted after the SPPF module and before the three-scale outputs in the neck,using cascaded channel and spatial attention to enhance the model's focus on fine-grained lesion textures and small targets.Experiments on a public dataset containing 10,852 images covering 10 common tomato diseases demonstrate that the proposed Ghost-CBAM model achieves an mAP@0.5∶0.95 of 0.8762,surpassing the original YOLOv8n baseline by 0.94 percentage points and the CBAM-only variant by 0.25 percentage points.Meanwhile,the model parameter count is reduced to only 1.73×106(42.5％smaller than the baseline),with a weight file of just 3.7 MB.The proposed method effectively balances detection accuracy and model lightweighting,offering a practical solution for edge deployment in field settings.