We succeed in inserting a number of nitrogen atoms into the R<sub>2</sub>Fe<sub>17</sub>and RTiFe<sub>11</sub>intermetalliccompounds.The nitrides retain original Th<sub>2</...We succeed in inserting a number of nitrogen atoms into the R<sub>2</sub>Fe<sub>17</sub>and RTiFe<sub>11</sub>intermetalliccompounds.The nitrides retain original Th<sub>2</sub>Zn<sub>17</sub>(or Th<sub>2</sub>Ni<sub>17<sub><</sub>/sub>and ThMn<sub>12</sub>structures,but with an increase inthe unit cell volume.The interstitial nitrogen atoms are found to have an effect of increasing the Curie tempera-ture and saturation magnetization.Moreover,a distinct effect on the magnetocrystalline anisotropy is also ob-served in these nitride compounds.Especially,NdTiFe<sub>11</sub>N<sub>0.5</sub>and SmTiFe<sub>11</sub>N<sub>2</sub> have excellent intrinsic magneticproperties favourable for permanent magnet applications.展开更多
The carbon and nitrogen atoms can be introduced into the R<sub>2</sub>Fe<sub>17</sub> structure by a proper heat treatment.The crystallographic and intrinsic magnetic properties of a series of ...The carbon and nitrogen atoms can be introduced into the R<sub>2</sub>Fe<sub>17</sub> structure by a proper heat treatment.The crystallographic and intrinsic magnetic properties of a series of R<sub>2</sub>Fe<sub>17</sub>CN<sub>x</sub>,where R=Y,Sm,Gd,Tb,Dy,Er have been investigated by using X-ray diffraction technique and magnetic measurements.The interstitial car-bon and nitrogen atoms are found to have an effect of dilating unit cell volume,raising Curie temperature andsaturation magnetization,as well as enhancing uniaxial magnetocrystalline anisotropy.展开更多
作为人工智能计算机视觉领域一项重要的任务,3D人体姿态估计受到了广泛的关注,并成功地应用在人机交互、电影游戏制作等领域。然而,3D人体姿态估计仍然面临着很大的挑战,主要是人体遮挡问题和数据集视角冗余问题,这些问题严重影响了3D...作为人工智能计算机视觉领域一项重要的任务,3D人体姿态估计受到了广泛的关注,并成功地应用在人机交互、电影游戏制作等领域。然而,3D人体姿态估计仍然面临着很大的挑战,主要是人体遮挡问题和数据集视角冗余问题,这些问题严重影响了3D人体姿态估计结果精度与速度的提升。本文提出了一种基于多特征提取的3D人体姿态估计方法。首先通过采集多个相机视角下的图片数据,将所采图片数据放入2D人体关节点检测网络模型中,得到人体2D关节点。接着将采集到的人体数据输入到关节点置信度计算网络模型,得到视角图片中各个关节点的权重值。随后将2D人体关节点热图通过一个热图权重计算网络计算出热图权重,将各个视角下的权重特征计算融合得到加权后的2D人体关节点热图。最后将所得加权后的2D人体关节点热图和视角图片中各个关节点的权重值输入到三角化算法中,映射得到空间中的3D人体关节点。本文的关键思想是设计一个关节点置信度计算网络从输入图像中学习每个关节的置信度权重,同时提取了反映热图特征质量的权重矩阵,以提高遮挡视图中热图的特征质量。此外,使用感知哈希算法对Occlusion-Person数据集进行去视角实验,在保证结果准确性的同时提高了模型推理速度。本文方法是端到端可微的,可以显著地提高算法效率和鲁棒性。本文在Human3.6M和Occlusion-Person两个公共数据集上使用平均关节位置误差(Mean Per Joint Position Error,MPJPE)指标对该方法进行评估,分别取得27.3 mm和9.7 mm的结果。实验结果表明,该算法与最先进的方法相比,性能有了显著提升。展开更多
文摘We succeed in inserting a number of nitrogen atoms into the R<sub>2</sub>Fe<sub>17</sub>and RTiFe<sub>11</sub>intermetalliccompounds.The nitrides retain original Th<sub>2</sub>Zn<sub>17</sub>(or Th<sub>2</sub>Ni<sub>17<sub><</sub>/sub>and ThMn<sub>12</sub>structures,but with an increase inthe unit cell volume.The interstitial nitrogen atoms are found to have an effect of increasing the Curie tempera-ture and saturation magnetization.Moreover,a distinct effect on the magnetocrystalline anisotropy is also ob-served in these nitride compounds.Especially,NdTiFe<sub>11</sub>N<sub>0.5</sub>and SmTiFe<sub>11</sub>N<sub>2</sub> have excellent intrinsic magneticproperties favourable for permanent magnet applications.
基金The project supported by the National Natural Science Foundation of Chinathe Open Magnetism Laboratory of the Chinese Academy Sciences
文摘The carbon and nitrogen atoms can be introduced into the R<sub>2</sub>Fe<sub>17</sub> structure by a proper heat treatment.The crystallographic and intrinsic magnetic properties of a series of R<sub>2</sub>Fe<sub>17</sub>CN<sub>x</sub>,where R=Y,Sm,Gd,Tb,Dy,Er have been investigated by using X-ray diffraction technique and magnetic measurements.The interstitial car-bon and nitrogen atoms are found to have an effect of dilating unit cell volume,raising Curie temperature andsaturation magnetization,as well as enhancing uniaxial magnetocrystalline anisotropy.
文摘作为人工智能计算机视觉领域一项重要的任务,3D人体姿态估计受到了广泛的关注,并成功地应用在人机交互、电影游戏制作等领域。然而,3D人体姿态估计仍然面临着很大的挑战,主要是人体遮挡问题和数据集视角冗余问题,这些问题严重影响了3D人体姿态估计结果精度与速度的提升。本文提出了一种基于多特征提取的3D人体姿态估计方法。首先通过采集多个相机视角下的图片数据,将所采图片数据放入2D人体关节点检测网络模型中,得到人体2D关节点。接着将采集到的人体数据输入到关节点置信度计算网络模型,得到视角图片中各个关节点的权重值。随后将2D人体关节点热图通过一个热图权重计算网络计算出热图权重,将各个视角下的权重特征计算融合得到加权后的2D人体关节点热图。最后将所得加权后的2D人体关节点热图和视角图片中各个关节点的权重值输入到三角化算法中,映射得到空间中的3D人体关节点。本文的关键思想是设计一个关节点置信度计算网络从输入图像中学习每个关节的置信度权重,同时提取了反映热图特征质量的权重矩阵,以提高遮挡视图中热图的特征质量。此外,使用感知哈希算法对Occlusion-Person数据集进行去视角实验,在保证结果准确性的同时提高了模型推理速度。本文方法是端到端可微的,可以显著地提高算法效率和鲁棒性。本文在Human3.6M和Occlusion-Person两个公共数据集上使用平均关节位置误差(Mean Per Joint Position Error,MPJPE)指标对该方法进行评估,分别取得27.3 mm和9.7 mm的结果。实验结果表明,该算法与最先进的方法相比,性能有了显著提升。
