The interfacial instability of the poly(ethylene oxide)(PEO)-based electrolytes impedes the long-term cycling and further application of all-solid-state lithium metal batter-ies.In this work,we have shown an effective...The interfacial instability of the poly(ethylene oxide)(PEO)-based electrolytes impedes the long-term cycling and further application of all-solid-state lithium metal batter-ies.In this work,we have shown an effective additive 1-adaman-tanecarbonitrile,which con-tributes to the excellent per-formance of the poly(ethylene oxide)-based electrolytes.Owing to the strong interaction of the 1-Adamantanecarboni-trile to the polymer matrix and anions,the coordination of the Li^(+)-EO is weakened,and the binding effect of anions is strengthened,thereby improving the Li^(+)conductivity and the electrochemical stability.The diamond building block on the surface of the lithium anode can sup-press the growth of lithium dendrites.Importantly,the 1-Adamantanecarbonitrile also regulates the formation of LiF in the solid electrolyte interface and cathode electrolyte interface,which contributes to the interfacial stability(especially at high voltages)and protects the electrodes,enabling all-solid-state batteries to cycle at high voltages for long periods of time.Therefore,the Li/Li symmetric cell undergoes long-term lithium plating/stripping for more than 2000 h.1-Adamantanecarbonitrile-poly(ethylene oxide)-based LFP/Li and 4.3 V Ni_(0.8)Mn_(0.1)Co_(0.1)O_(2)/Li all-solid-state batteries achieved stable cycles for 1000 times,with capacity retention rates reaching 85%and 80%,respectively.展开更多
目前主流人体动作识别大部分都是基于卷积神经网络(Convolutional Neural Network,CNN)实现,而CNN容易忽略视频中的空间位置信息,从而降低了视频空间频域中动作识别能力。同时传统CNN不能快速定位到关键的特征位置,并且在训练过程中不...目前主流人体动作识别大部分都是基于卷积神经网络(Convolutional Neural Network,CNN)实现,而CNN容易忽略视频中的空间位置信息,从而降低了视频空间频域中动作识别能力。同时传统CNN不能快速定位到关键的特征位置,并且在训练过程中不能并行计算导致效率低。为了解决传统CNN在处理时间频域和多并行计算问题,提出了基于视觉Transformer(Vision Transformer,ViT)和3D卷积网络学习时空特征(Learning Spatiotemporal Features with 3D Convolutional Network,C3D)的人体动作识别算法。使用C3D提取视频的多维特征图、ViT的特征切片窗口对多维特征进行全局特征分割;使用Transformer的编码-解码模块对视频中人体动作进行预测。实验结果表明,所提的人体动作识别算法在UCF-101、HMDB51数据集上提高了动作识别的准确率。展开更多
基金supported by National Natural Science Foundation of China(Grant No.22209012).
文摘The interfacial instability of the poly(ethylene oxide)(PEO)-based electrolytes impedes the long-term cycling and further application of all-solid-state lithium metal batter-ies.In this work,we have shown an effective additive 1-adaman-tanecarbonitrile,which con-tributes to the excellent per-formance of the poly(ethylene oxide)-based electrolytes.Owing to the strong interaction of the 1-Adamantanecarboni-trile to the polymer matrix and anions,the coordination of the Li^(+)-EO is weakened,and the binding effect of anions is strengthened,thereby improving the Li^(+)conductivity and the electrochemical stability.The diamond building block on the surface of the lithium anode can sup-press the growth of lithium dendrites.Importantly,the 1-Adamantanecarbonitrile also regulates the formation of LiF in the solid electrolyte interface and cathode electrolyte interface,which contributes to the interfacial stability(especially at high voltages)and protects the electrodes,enabling all-solid-state batteries to cycle at high voltages for long periods of time.Therefore,the Li/Li symmetric cell undergoes long-term lithium plating/stripping for more than 2000 h.1-Adamantanecarbonitrile-poly(ethylene oxide)-based LFP/Li and 4.3 V Ni_(0.8)Mn_(0.1)Co_(0.1)O_(2)/Li all-solid-state batteries achieved stable cycles for 1000 times,with capacity retention rates reaching 85%and 80%,respectively.
文摘目前主流人体动作识别大部分都是基于卷积神经网络(Convolutional Neural Network,CNN)实现,而CNN容易忽略视频中的空间位置信息,从而降低了视频空间频域中动作识别能力。同时传统CNN不能快速定位到关键的特征位置,并且在训练过程中不能并行计算导致效率低。为了解决传统CNN在处理时间频域和多并行计算问题,提出了基于视觉Transformer(Vision Transformer,ViT)和3D卷积网络学习时空特征(Learning Spatiotemporal Features with 3D Convolutional Network,C3D)的人体动作识别算法。使用C3D提取视频的多维特征图、ViT的特征切片窗口对多维特征进行全局特征分割;使用Transformer的编码-解码模块对视频中人体动作进行预测。实验结果表明,所提的人体动作识别算法在UCF-101、HMDB51数据集上提高了动作识别的准确率。
