针对当前药品名称检测存在检测精度低、算法参数量大和现有公开数据集在小目标检测上存在限制等问题,并基于YOLOv8算法,提出了一种结合Slim-Neck替换特征融合层、GAM注意力机制的检测精度高、硬件要求低的新模型YOLOv8-SNG.使用Drug Name Detection Dataset数据集验证算法的有效性.实验结果表明,与YOLOv8相比,在Drug Name Detection Dataset数据集上FPS少量上涨的情况下,mAP@0.5提高了11.9％,mAP@0.5:0.95提高了6.43％;改进后的算法具有更强的泛化性能,能够更好的适用于药品名称检测任务.
Drug Name Detection Algorithm Based on Improved YOLOv8
Aiming at the problems of low detection accuracy,large number of algorithm parameters and limitations of existing public data sets in small target detection,a new model YOLOv8-SNG was proposed based on the YOLOv8 algorithm,which combined Slim-Neck replacement feature fusion layer,GAM attention mechanism with high detection accuracy and low hardware requirements.The Drug Name Detection Dataset was used to verify the validity of the algorithm.The experimental results show that,compared with YOLOv8,when FPS in the Drug Name Detection Dataset slightly increased,mAP@0.5 increased by 11.9％and mAP@0.5:0.95 increased by 6.43％.The improved algorithm has stronger generalization performance and can be better applied to the drug name detection task.
the name of the drugtarget detectionattention mechanismfeature fusion layerk
万方数据
维普
