针对无人机图像中由于目标微小且相互遮挡、特征信息少导致检测精度低的问题,提出一种基于改进YOLOv7的无人机图像目标检测算法。在颈部和检测头中加入了坐标卷积,能更好地感受特征图中目标的位置信息;增加P2检测层,减少小目标特征丢失...针对无人机图像中由于目标微小且相互遮挡、特征信息少导致检测精度低的问题,提出一种基于改进YOLOv7的无人机图像目标检测算法。在颈部和检测头中加入了坐标卷积,能更好地感受特征图中目标的位置信息;增加P2检测层,减少小目标特征丢失、提高小目标检测能力;提出多信息流融合注意力机制——Spatial and Channel Attention Mechanism(SCA),动态调整注意力对空间信息流和语义信息流的关注,获得更丰富的特征信息以提高捕获目标的能力;更换损失函数为SIoU,加快模型收敛速度。在公开数据集VisDrone2019上进行对比实验,改进后算法的mAP50值相比YOLOv7提高了4%,达到了52.4%,FPS为37,消融实验验证了每个模块均提升了检测精度。实验表明,改进后的算法能较好地检测无人机图像中的目标。展开更多
Electrochemical reduction of CO_(2) has attracted wide attention because it can realize an artificial carbon cycle.The key step to achieving efficient electrochemical conversion of CO_(2) is the preparation of catalys...Electrochemical reduction of CO_(2) has attracted wide attention because it can realize an artificial carbon cycle.The key step to achieving efficient electrochemical conversion of CO_(2) is the preparation of catalysts with multiple active sites,prepared without additional strong alkaline solutions.In this work,the surfaces of Ag foils were nanosized by simple cyclic voltammetry(CV)pretreatment,which can supply an abundant specific surface area for the CO_(2) reduction reaction(CO_(2)RR).The results indicated that the CO_(2)RR performance of Ag catalysts was boosted,and the competitive reaction of H_(2) evolution was suppressed.For this catalyst,at-0.92 V vs.RHE(reversible hydrogen electrode),the current density of CO achieved 12.31 mA/cm^(2),a significant increase compared to that of the untreated Ag foils(0.81 mA/cm^(2)).The outstanding electrochemical properties are ascribed to the rich nanostructures on Ag foils after the CV pretreatment.The nanostructure has a very high specific surface area,so it can supply more reactive active sites in the process of CO_(2)RR.展开更多
文摘针对无人机图像中由于目标微小且相互遮挡、特征信息少导致检测精度低的问题,提出一种基于改进YOLOv7的无人机图像目标检测算法。在颈部和检测头中加入了坐标卷积,能更好地感受特征图中目标的位置信息;增加P2检测层,减少小目标特征丢失、提高小目标检测能力;提出多信息流融合注意力机制——Spatial and Channel Attention Mechanism(SCA),动态调整注意力对空间信息流和语义信息流的关注,获得更丰富的特征信息以提高捕获目标的能力;更换损失函数为SIoU,加快模型收敛速度。在公开数据集VisDrone2019上进行对比实验,改进后算法的mAP50值相比YOLOv7提高了4%,达到了52.4%,FPS为37,消融实验验证了每个模块均提升了检测精度。实验表明,改进后的算法能较好地检测无人机图像中的目标。
基金supported by the National Natural Science Foundation of China(No.22179016)。
文摘Electrochemical reduction of CO_(2) has attracted wide attention because it can realize an artificial carbon cycle.The key step to achieving efficient electrochemical conversion of CO_(2) is the preparation of catalysts with multiple active sites,prepared without additional strong alkaline solutions.In this work,the surfaces of Ag foils were nanosized by simple cyclic voltammetry(CV)pretreatment,which can supply an abundant specific surface area for the CO_(2) reduction reaction(CO_(2)RR).The results indicated that the CO_(2)RR performance of Ag catalysts was boosted,and the competitive reaction of H_(2) evolution was suppressed.For this catalyst,at-0.92 V vs.RHE(reversible hydrogen electrode),the current density of CO achieved 12.31 mA/cm^(2),a significant increase compared to that of the untreated Ag foils(0.81 mA/cm^(2)).The outstanding electrochemical properties are ascribed to the rich nanostructures on Ag foils after the CV pretreatment.The nanostructure has a very high specific surface area,so it can supply more reactive active sites in the process of CO_(2)RR.
