The Shanghai Cooperation Organization (SCO) Management System Certification Research Class formally opened in Beijing on September 6, 2007. The research class was attended by government officials and experts in th... The Shanghai Cooperation Organization (SCO) Management System Certification Research Class formally opened in Beijing on September 6, 2007. The research class was attended by government officials and experts in the fields of industry, commerce,environmental protection,statistics, standardization,metrology, detection and certification from Kazakhstan, Kyrgyzstan, Uzbekistan and Tajikistan.……展开更多
Class III malocclusion associated with skeletal anterior open bite pattern in adults can be a challenging orthodontic problem, especially for the nonsurgical treatment. Conventionally, several treatment alternatives a...Class III malocclusion associated with skeletal anterior open bite pattern in adults can be a challenging orthodontic problem, especially for the nonsurgical treatment. Conventionally, several treatment alternatives are available such as tooth extraction, molar intrusion, and absolute anchorage system or orthognathic surgical correction. Although correction with surgery may be the most effective and stable way, many patients refused surgical treatment plan because of the costs and traumas it may bring. We reported a nonsurgical orthopedic treatment of 22-year-old male with severe skeletal anterior open bite, dental Class III malocclusion, posterior crossbite and a high mandibular plane angle. The patient refused surgery and extraction. So we formulated a treatment plan consisting of using rapid palatal expansion appliance to expand the maxilla, standard edgewise brackets to align the teeth, Class III elastics to correct the canines, premolars, and molars relationship, reverse curve of the nickel-titanium wire combined with anterior vertical elastics to intrude molars and correct open bite. In this case, without suffering of surgery, the posterior crossbite was ideally corrected, and ideal overjet and overbite relationships and functional occlusion were all achieved. The patient obtained satisfactory occlusal as well as functional and stable results.展开更多
开放世界目标检测(open world object detection,OWOD)的主要任务是检测已知类别和识别未知目标。此外,模型在下一个训练阶段中逐步学习已知新类。针对OW-DETR(open-world detection transformer)中未知类召回率偏低、密集目标与小目标...开放世界目标检测(open world object detection,OWOD)的主要任务是检测已知类别和识别未知目标。此外,模型在下一个训练阶段中逐步学习已知新类。针对OW-DETR(open-world detection transformer)中未知类召回率偏低、密集目标与小目标漏检等问题,提出了一种UBA-OWDT(UCSO,BiStrip and AFDF of open-world detection transformer)开放世界目标检测网络。针对未知类召回率偏低的问题,对未知类评分优化(unknown class scoring optimization,UCSO),将生成的浅层类激活图与聚合类激活图融合,获取细粒度特征信息,提高未知类的目标评分,进而提升未知类的召回率;针对小目标漏检问题,将双条状注意力(spatial attention based on strip pooling and strip convolution,BiStrip)应用于输入特征图,捕获长程依赖,保留目标精确的位置信息,增强感兴趣目标的表征,以检测小目标;针对密集目标漏检问题,采用自适应特征动态融合(adaptive feature dynamic fusion,AFDF),根据目标大小和形状,获得不同的感受野,动态分配注意力权重,关注目标的重要部分,融合不同层级的特征,以检测密集目标。在OWOD数据集的实验结果表明,未知类召回率增值范围在0.7~1.5个百分点,mAP增值范围在0.6~1.2个百分点,与现有的开放世界目标检测方法相比,在召回率偏低、密集目标与小目标漏检问题上具有一定的优势。展开更多
开集分类识别是近10多年来模式识别领域研究的热点,它能够识别训练集中已知类别的测试样本,同时还能够有效“拒识”未知类别的测试样本;这些未知类别样本不包含在训练集中。现有的开集分类识别算法主要是基于Support Vector Machine(SVM...开集分类识别是近10多年来模式识别领域研究的热点,它能够识别训练集中已知类别的测试样本,同时还能够有效“拒识”未知类别的测试样本;这些未知类别样本不包含在训练集中。现有的开集分类识别算法主要是基于Support Vector Machine(SVM)和深度学习网络框架进行改进,并且主要应用在自然景物图像领域中;在光谱分析领域中还鲜有报道。将传统的闭集框架下的模糊推理分类器进行模型改进,提出了开集框架下的改进模糊推理分类器,并将其应用到木材树种近红外光谱分类识别中。首先,使用Flame-NIR近红外微型光谱仪采集木材样本横切面的近红外光谱曲线,采用Metric Learning算法进行光谱向量维度约简降维至4维(4D)。其次,改进闭集框架下的模糊推理分类器,根据模糊规则置信度和各维度隶属度概率的乘积构建Generalized Basic Probability Assignment(GBPA),再根据GBPA进行分类处理。在20个树种的具有不同的Openness指标下的近红外光谱数据集的分类识别对比实验表明,改进的开集模糊推理分类器(fuzzy reasoning classifier in an open set,FRCOS)优于现有的基于机器学习和深度学习的开集分类识别主流算法,具有较好的评价指标F-Score,Kappa系数及总体识别率。展开更多
文摘 The Shanghai Cooperation Organization (SCO) Management System Certification Research Class formally opened in Beijing on September 6, 2007. The research class was attended by government officials and experts in the fields of industry, commerce,environmental protection,statistics, standardization,metrology, detection and certification from Kazakhstan, Kyrgyzstan, Uzbekistan and Tajikistan.……
文摘Class III malocclusion associated with skeletal anterior open bite pattern in adults can be a challenging orthodontic problem, especially for the nonsurgical treatment. Conventionally, several treatment alternatives are available such as tooth extraction, molar intrusion, and absolute anchorage system or orthognathic surgical correction. Although correction with surgery may be the most effective and stable way, many patients refused surgical treatment plan because of the costs and traumas it may bring. We reported a nonsurgical orthopedic treatment of 22-year-old male with severe skeletal anterior open bite, dental Class III malocclusion, posterior crossbite and a high mandibular plane angle. The patient refused surgery and extraction. So we formulated a treatment plan consisting of using rapid palatal expansion appliance to expand the maxilla, standard edgewise brackets to align the teeth, Class III elastics to correct the canines, premolars, and molars relationship, reverse curve of the nickel-titanium wire combined with anterior vertical elastics to intrude molars and correct open bite. In this case, without suffering of surgery, the posterior crossbite was ideally corrected, and ideal overjet and overbite relationships and functional occlusion were all achieved. The patient obtained satisfactory occlusal as well as functional and stable results.
文摘开放世界目标检测(open world object detection,OWOD)的主要任务是检测已知类别和识别未知目标。此外,模型在下一个训练阶段中逐步学习已知新类。针对OW-DETR(open-world detection transformer)中未知类召回率偏低、密集目标与小目标漏检等问题,提出了一种UBA-OWDT(UCSO,BiStrip and AFDF of open-world detection transformer)开放世界目标检测网络。针对未知类召回率偏低的问题,对未知类评分优化(unknown class scoring optimization,UCSO),将生成的浅层类激活图与聚合类激活图融合,获取细粒度特征信息,提高未知类的目标评分,进而提升未知类的召回率;针对小目标漏检问题,将双条状注意力(spatial attention based on strip pooling and strip convolution,BiStrip)应用于输入特征图,捕获长程依赖,保留目标精确的位置信息,增强感兴趣目标的表征,以检测小目标;针对密集目标漏检问题,采用自适应特征动态融合(adaptive feature dynamic fusion,AFDF),根据目标大小和形状,获得不同的感受野,动态分配注意力权重,关注目标的重要部分,融合不同层级的特征,以检测密集目标。在OWOD数据集的实验结果表明,未知类召回率增值范围在0.7~1.5个百分点,mAP增值范围在0.6~1.2个百分点,与现有的开放世界目标检测方法相比,在召回率偏低、密集目标与小目标漏检问题上具有一定的优势。
文摘开集分类识别是近10多年来模式识别领域研究的热点,它能够识别训练集中已知类别的测试样本,同时还能够有效“拒识”未知类别的测试样本;这些未知类别样本不包含在训练集中。现有的开集分类识别算法主要是基于Support Vector Machine(SVM)和深度学习网络框架进行改进,并且主要应用在自然景物图像领域中;在光谱分析领域中还鲜有报道。将传统的闭集框架下的模糊推理分类器进行模型改进,提出了开集框架下的改进模糊推理分类器,并将其应用到木材树种近红外光谱分类识别中。首先,使用Flame-NIR近红外微型光谱仪采集木材样本横切面的近红外光谱曲线,采用Metric Learning算法进行光谱向量维度约简降维至4维(4D)。其次,改进闭集框架下的模糊推理分类器,根据模糊规则置信度和各维度隶属度概率的乘积构建Generalized Basic Probability Assignment(GBPA),再根据GBPA进行分类处理。在20个树种的具有不同的Openness指标下的近红外光谱数据集的分类识别对比实验表明,改进的开集模糊推理分类器(fuzzy reasoning classifier in an open set,FRCOS)优于现有的基于机器学习和深度学习的开集分类识别主流算法,具有较好的评价指标F-Score,Kappa系数及总体识别率。
