ID 原文 译文
653 其次，本文又提出了一种能够突显整段行为视频序列中较为重要帧的视频帧关注模型，以减少异类行为视频序列间相似的前后关联给识别带来的干扰。 Secondly, this paper proposes a video frame attention model (VFA)that can highlight the more impor-tant frames in the video sequence of the whole action, so as to reduce the interference brought by the similar before and after correlation between video sequences of different actions.
654 最后，提出了一个能够端到端训练的网络模型:基于循环区域关注和视频帧关注的视频行为识别网络(Recurrent Region Attention andVideo Frame Attention based video action recognition Network，RFANet)。 Finally, this paper presents a network model which can performend-to-end training: recurrent region attention and video frame attention based video action recognition network (RFANet).
655 在两个视频行为识别基准 UCF101 数据集和 HM-DB51 数据集上的实验表明，本文提出的端到端网络 RFANet 能够可靠地识别出视频中行为的所属类别。 Experiments on two video action recognition benchmark UCF101 dataset and HMDB51 dataset show that the RFANet pro-posed in this paper can reliably identify the category of action in the video.
656 受双流结构启发，本文构建了双模态 RFANet 网络。 Inspired by the two-stream structure, we constructa two-modalities RFANet network.
657 在相同的训练环境下，双模态 RFANet 网络在两个数据集上达到了最优的性能。 In the same training conditions, the two-modalities RFANet network achieved optimal performance on both datasets.
658 针对卫星磁特性研究与磁试验，建立了卫星多磁偶极子模型，分离了线性和非线性部分。 According to the research of satellite magnetic characteristics and magnetic test, the model of satellite multi-magnetic dipole is established, the linear part and nonlinear parts are separated.
659 利用遗传算法，以测点的模拟值和测量值之差的均方根作为目标函数，搜索多磁偶极子的位置，同时检测拟合误差，使得目标函数达到最优的同时，拟合误差也满足要求。 The root mean square of the difference between the simulated value and the measured value is taken as the objective function, and the genetic algorithm is used to search the position of the multi-magnetic dipoles, at the same time, the fitness error between measurements and calculations is monitored so that when the object function reaches the optimum, the fitness error can meet the requirement.
660 最后对卫星静止状态和加电状态的磁性进行了计算，试验结果表明，用该方法建立卫星磁场模型精度高，磁矩和磁场的误差均小于 5% ，并且对测量误差不敏感，具有很强的鲁棒性。 At last, the magnetic properties of the satellite in static state and electrified state are calculated, the result shows that the magnetic model obtained by adopting the method is high accurate and strong robust, the error of magnetic field and magnetic moment areboth less than 5% .
661 针对深度神经网络模型仅学习当前指代链语义信息忽略了单个指代链识别结果的长期影响问题，提出一种结合深度强化学习(deep reinforcement learning)的维吾尔语人称代词指代消解方法。 Deep neural network models for Uyghur personal pronouns resolution learn semantic information for cur-rent anaphora chain, but ignore the long-term effects of single anaphora chain recognition results. This paper proposes aUyghur personal pronoun anaphora resolution based on deep reinforcement learning.
662 该方法将指代消解任务定义为强化学习环境下顺序决策过程，有效利用之前状态中先行语信息判定当前指代链指代关系。 This method defines the anaphora res-olution task as the sequential decision process under the reinforcement learning environment, and effectively uses the ante-cedent information in the previous state to determine the current personal pronoun-candidate antecedent pairs.