To solve the problem of low detection accuracy for complex weld defects,the paper proposes a weld defects detection method based on improved YOLOv5s.To enhance the ability to focus on key information in feature maps,t...To solve the problem of low detection accuracy for complex weld defects,the paper proposes a weld defects detection method based on improved YOLOv5s.To enhance the ability to focus on key information in feature maps,the scSE attention mechanism is intro-duced into the backbone network of YOLOv5s.A Fusion-Block module and additional layers are added to the neck network of YOLOv5s to improve the effect of feature fusion,which is to meet the needs of complex object detection.To reduce the computation-al complexity of the model,the C3Ghost module is used to replace the CSP2_1 module in the neck network of YOLOv5s.The scSE-ASFF module is constructed and inserted between the neck network and the prediction end,which is to realize the fusion of features between the different layers.To address the issue of imbalanced sample quality in the dataset and improve the regression speed and accuracy of the loss function,the CIoU loss function in the YOLOv5s model is replaced with the Focal-EIoU loss function.Finally,ex-periments are conducted based on the collected weld defect dataset to verify the feasibility of the improved YOLOv5s for weld defects detection.The experimental results show that the precision and mAP of the improved YOLOv5s in detecting complex weld defects are as high as 83.4%and 76.1%,respectively,which are 2.5%and 7.6%higher than the traditional YOLOv5s model.The proposed weld defects detection method based on the improved YOLOv5s in this paper can effectively solve the problem of low weld defects detection accuracy.展开更多
In recent years,the problem of privacy leakage has attracted increasing attentions.Therefore,machine learning privacy protection becomes crucial research topic.In this paper,the Paillier homomorphic encryption algorit...In recent years,the problem of privacy leakage has attracted increasing attentions.Therefore,machine learning privacy protection becomes crucial research topic.In this paper,the Paillier homomorphic encryption algorithm is proposed to protect the privacy data.The original LeNet-5 convolutional neural network model was first improved.Then the activation function was modified and the C5 layer was removed to reduce the number of model parameters and improve the operation efficiency.Finally,by mapping the operation of each layer in the convolutional neural network from the plaintext domain to the ciphertext domain,an improved LeNet-5 model that can run on encrypted data was constructed.The purpose of using machine learning algorithmwas realized and privacywas ensured at the same time.The analysis shows that the model is feasible and the efficiency is improved.展开更多
Glaucoma is characterized by the progressive loss of retinal ganglion cells (RGCs),although the pathogenic mechanism remains largely unknown.To study the mechanism and assess RGC degradation,mouse models are often use...Glaucoma is characterized by the progressive loss of retinal ganglion cells (RGCs),although the pathogenic mechanism remains largely unknown.To study the mechanism and assess RGC degradation,mouse models are often used to simulate human glaucoma and specific markers are used to label and quantify RGCs.However,manually counting RGCs is time-consuming and prone to distortion due to subjective bias.Furthermore,semi-automated counting methods can produce significant differences due to different parameters,thereby failing objective evaluation.Here,to improve counting accuracy and efficiency,we developed an automated algorithm based on the improved YOLOv5 model,which uses five channels instead of one,with a squeeze-and-excitation block added.The complete number of RGCs in an intact mouse retina was obtained by dividing the retina into small overlapping areas and counting,and then merging the divided areas using a non-maximum suppression algorithm.The automated quantification results showed very strong correlation (mean Pearson correlation coefficient of 0.993) with manual counting.Importantly,the model achieved an average precision of 0.981.Furthermore,the graphics processing unit (GPU) calculation time for each retina was less than 1 min.The developed software has been uploaded online as a free and convenient tool for studies using mouse models of glaucoma,which should help elucidate disease pathogenesis and potential therapeutics.展开更多
The dehydroaramatization of methane over W-supported ZSM-5 with varying degrees of Li+ ion-exchanged catalysts was studied with and without oxygen at 1073 K and atmospheric pressure. Catalyst activity and stability we...The dehydroaramatization of methane over W-supported ZSM-5 with varying degrees of Li+ ion-exchanged catalysts was studied with and without oxygen at 1073 K and atmospheric pressure. Catalyst activity and stability were found to be influenced by the catalyst acidity related to Bronsted acid sites and by the presence of oxygen in the feed. The NH3-TPD and FTIR-pyridine results demonstrated that partially exchanged of H+ ions by Li+ into the W/HZSM-5 catalysts could be used to control the amount of strong acid sites on the catalyst surface. Without oxygen, the 3WHLi-Z (5:1) catalyst that has strong acid sites equal to nearly 74% of the original strong acid sites in the parent HZSM-5 exhibited the highest methane conversion and selectivity towards aromatics. However, the catalyst deactivated in a five hour period. In the presence of oxygen, the catalyst activity and stability could be improved further. The results of this study revealed that a suitable amount of strong Bronsted acid sites as well as oxygen addition in the feed increased the catalyst activity and stability. The 3WHLi-Z(5:1) catalyst exhibited improved performance in the dehydroaromatization of methane.展开更多
The statistics of the number of rapeseed seedlings are very important for breeders and planters to conduct seed quality testing,field crop management and yield estimation.Calculating the number of seedlings is ineffic...The statistics of the number of rapeseed seedlings are very important for breeders and planters to conduct seed quality testing,field crop management and yield estimation.Calculating the number of seedlings is inefficient and cumbersome in the traditional method.In this study,a method was proposed for efficient detection and calculation of rapeseed seedling number based on improved you only look once version 5(YOLOv5)to identify objects and deep-sort to perform object tracking for rapeseed seedling video.Coordinated attention(CA)mechanism was added to the trunk of the improved YOLOv5s,which made the model more effective in identifying shaded,dense and small rapeseed seedlings.Also,the use of the GSConv module replaced the standard convolution at the neck,reduced model parameters and enabled it better able to be equipped for mobile devices.The accuracy and recall rate of using improved YOLOv5s on the test set by 1.9%and 3.7%compared to 96.2%and 93.7%of YOLOv5s,respectively.The experimental results showed that the average error of monitoring the number of seedlings by unmanned aerial vehicles(UAV)video of rapeseed seedlings based on improved YOLOv5s combined with depth-sort method was 4.3%.The presented approach can realize rapid statistics of the number of rapeseed seedlings in the field based on UAV remote sensing,provide a reference for variety selection and precise management of rapeseed.展开更多
There is distinct difference in the tangential wind profile between different typhoons in the western North Pacific. At present, only two parameters, maximum wind and radius of maximum wind, are used in NCAR-AFWA bogu...There is distinct difference in the tangential wind profile between different typhoons in the western North Pacific. At present, only two parameters, maximum wind and radius of maximum wind, are used in NCAR-AFWA bogus for MM5 mesoscale numerical model. As a result, sometimes the outer structure of typhoon cannot be described accurately. The tangential wind profile of NCAR-AFWA bogus is improved by introducing radii of 25.7 m/s and 15.4 m/s, and then the track and intensity of Typhoon Nockten (No.0425) are simulated. The results show that the simulations of track and intensity of typhoon both have been improved by simultaneously introducing the radii in the tangential wind profile of typhoon bogus. At the same time, there is improvement in the gale wind range of the typhoon simulated.展开更多
引入了Inception-SE卷积模块组来提升LeNet-5网络的广度与深度,运用SE模块增强了有用的特征并抑制了对当前任务用处不大的特征;使用BN层和Dropout优化网络,防止梯度弥散,提升精度;使用全局池化层(global average pooling,GAP)代替全连...引入了Inception-SE卷积模块组来提升LeNet-5网络的广度与深度,运用SE模块增强了有用的特征并抑制了对当前任务用处不大的特征;使用BN层和Dropout优化网络,防止梯度弥散,提升精度;使用全局池化层(global average pooling,GAP)代替全连接层来减少网络计算参数.研究结果表明:改进后网络的识别精度达到了99.88%,比传统的LeNet-5网络提高了1.71%.展开更多
基金supported by Postgraduate Research&Practice Innovation Program of Jiangsu Province(Grant No.KYCX24_4084).
文摘To solve the problem of low detection accuracy for complex weld defects,the paper proposes a weld defects detection method based on improved YOLOv5s.To enhance the ability to focus on key information in feature maps,the scSE attention mechanism is intro-duced into the backbone network of YOLOv5s.A Fusion-Block module and additional layers are added to the neck network of YOLOv5s to improve the effect of feature fusion,which is to meet the needs of complex object detection.To reduce the computation-al complexity of the model,the C3Ghost module is used to replace the CSP2_1 module in the neck network of YOLOv5s.The scSE-ASFF module is constructed and inserted between the neck network and the prediction end,which is to realize the fusion of features between the different layers.To address the issue of imbalanced sample quality in the dataset and improve the regression speed and accuracy of the loss function,the CIoU loss function in the YOLOv5s model is replaced with the Focal-EIoU loss function.Finally,ex-periments are conducted based on the collected weld defect dataset to verify the feasibility of the improved YOLOv5s for weld defects detection.The experimental results show that the precision and mAP of the improved YOLOv5s in detecting complex weld defects are as high as 83.4%and 76.1%,respectively,which are 2.5%and 7.6%higher than the traditional YOLOv5s model.The proposed weld defects detection method based on the improved YOLOv5s in this paper can effectively solve the problem of low weld defects detection accuracy.
基金The National Natural Science Foundation of China(No.61572521)Engineering University of PAP Innovation Team Science Foundation(No.KYTD201805)Natural Science Basic Research Plan in Shaanxi Province of China(2021JM252).
文摘In recent years,the problem of privacy leakage has attracted increasing attentions.Therefore,machine learning privacy protection becomes crucial research topic.In this paper,the Paillier homomorphic encryption algorithm is proposed to protect the privacy data.The original LeNet-5 convolutional neural network model was first improved.Then the activation function was modified and the C5 layer was removed to reduce the number of model parameters and improve the operation efficiency.Finally,by mapping the operation of each layer in the convolutional neural network from the plaintext domain to the ciphertext domain,an improved LeNet-5 model that can run on encrypted data was constructed.The purpose of using machine learning algorithmwas realized and privacywas ensured at the same time.The analysis shows that the model is feasible and the efficiency is improved.
基金supported by the National Natural Science Foundation of China (61405028 to J.Z.,81770935 to H.B.Z.)Fundamental Research Funds for the Central Universities (University of Electronic Science and Technology of China)(ZYGX2019J053 to J.Z.)Department of Science and Technology of Sichuan Province,China (2020YJ0445 to H.B.Z.)。
文摘Glaucoma is characterized by the progressive loss of retinal ganglion cells (RGCs),although the pathogenic mechanism remains largely unknown.To study the mechanism and assess RGC degradation,mouse models are often used to simulate human glaucoma and specific markers are used to label and quantify RGCs.However,manually counting RGCs is time-consuming and prone to distortion due to subjective bias.Furthermore,semi-automated counting methods can produce significant differences due to different parameters,thereby failing objective evaluation.Here,to improve counting accuracy and efficiency,we developed an automated algorithm based on the improved YOLOv5 model,which uses five channels instead of one,with a squeeze-and-excitation block added.The complete number of RGCs in an intact mouse retina was obtained by dividing the retina into small overlapping areas and counting,and then merging the divided areas using a non-maximum suppression algorithm.The automated quantification results showed very strong correlation (mean Pearson correlation coefficient of 0.993) with manual counting.Importantly,the model achieved an average precision of 0.981.Furthermore,the graphics processing unit (GPU) calculation time for each retina was less than 1 min.The developed software has been uploaded online as a free and convenient tool for studies using mouse models of glaucoma,which should help elucidate disease pathogenesis and potential therapeutics.
文摘The dehydroaramatization of methane over W-supported ZSM-5 with varying degrees of Li+ ion-exchanged catalysts was studied with and without oxygen at 1073 K and atmospheric pressure. Catalyst activity and stability were found to be influenced by the catalyst acidity related to Bronsted acid sites and by the presence of oxygen in the feed. The NH3-TPD and FTIR-pyridine results demonstrated that partially exchanged of H+ ions by Li+ into the W/HZSM-5 catalysts could be used to control the amount of strong acid sites on the catalyst surface. Without oxygen, the 3WHLi-Z (5:1) catalyst that has strong acid sites equal to nearly 74% of the original strong acid sites in the parent HZSM-5 exhibited the highest methane conversion and selectivity towards aromatics. However, the catalyst deactivated in a five hour period. In the presence of oxygen, the catalyst activity and stability could be improved further. The results of this study revealed that a suitable amount of strong Bronsted acid sites as well as oxygen addition in the feed increased the catalyst activity and stability. The 3WHLi-Z(5:1) catalyst exhibited improved performance in the dehydroaromatization of methane.
文摘The statistics of the number of rapeseed seedlings are very important for breeders and planters to conduct seed quality testing,field crop management and yield estimation.Calculating the number of seedlings is inefficient and cumbersome in the traditional method.In this study,a method was proposed for efficient detection and calculation of rapeseed seedling number based on improved you only look once version 5(YOLOv5)to identify objects and deep-sort to perform object tracking for rapeseed seedling video.Coordinated attention(CA)mechanism was added to the trunk of the improved YOLOv5s,which made the model more effective in identifying shaded,dense and small rapeseed seedlings.Also,the use of the GSConv module replaced the standard convolution at the neck,reduced model parameters and enabled it better able to be equipped for mobile devices.The accuracy and recall rate of using improved YOLOv5s on the test set by 1.9%and 3.7%compared to 96.2%and 93.7%of YOLOv5s,respectively.The experimental results showed that the average error of monitoring the number of seedlings by unmanned aerial vehicles(UAV)video of rapeseed seedlings based on improved YOLOv5s combined with depth-sort method was 4.3%.The presented approach can realize rapid statistics of the number of rapeseed seedlings in the field based on UAV remote sensing,provide a reference for variety selection and precise management of rapeseed.
基金Experiments on coupling typhoons, waves and storm surges in the South China Sea andpreviews of typhoon-inflicted disasters, a project from the research foundation for the science of tropical andmarine meteorology
文摘There is distinct difference in the tangential wind profile between different typhoons in the western North Pacific. At present, only two parameters, maximum wind and radius of maximum wind, are used in NCAR-AFWA bogus for MM5 mesoscale numerical model. As a result, sometimes the outer structure of typhoon cannot be described accurately. The tangential wind profile of NCAR-AFWA bogus is improved by introducing radii of 25.7 m/s and 15.4 m/s, and then the track and intensity of Typhoon Nockten (No.0425) are simulated. The results show that the simulations of track and intensity of typhoon both have been improved by simultaneously introducing the radii in the tangential wind profile of typhoon bogus. At the same time, there is improvement in the gale wind range of the typhoon simulated.
文摘引入了Inception-SE卷积模块组来提升LeNet-5网络的广度与深度,运用SE模块增强了有用的特征并抑制了对当前任务用处不大的特征;使用BN层和Dropout优化网络,防止梯度弥散,提升精度;使用全局池化层(global average pooling,GAP)代替全连接层来减少网络计算参数.研究结果表明:改进后网络的识别精度达到了99.88%,比传统的LeNet-5网络提高了1.71%.
