ID 原文 译文
52997 结果表明，提出的算法在对目标进行有效检测的同时，定位效果比Faster RCNN更加精确，实现了边界框的精准回归。 The results show that the proposed algorithm is more accurate than the Faster RCNN in the effective detection of the target, and achieves the accurate regression of the bounding box.
52998 在图像分类中，图像标签的获取是昂贵的和费时的。 In image classification, the acquisition of image labels is expensive and time-consuming.
52999 为了减少标注成本，提出了一种主动半监督深度学习的归纳一致性预测算法(ICP-ASSDL)，该算法使用一种新颖的奇异值度量来产生可靠的置信度。 In order to reduce the labeling cost, an inductive conformal prediction for active semi-supervised deep learning(ICP-ASSDL) is proposed, which uses a novel singular value metric to produce reliability.
53000 ICP-ASSDL用4个标准(信息质量、边缘抽样、多样性和面向类别度量)从未标记池中选取实例来提高分类性能。 ICP-ASSDL uses four criteria(information quality, edge sampling, diversity, and category-oriented metrics) to select instances from an unlabeled pool to improve classification performance.
53001 面向类别度量采用蒙德里安一致性预测算法来减弱非平衡问题的影响。 For category-oriented metrics, the Mondrian consensus prediction algorithm is used to reduce the impact of non-equilibrium problems.
53002 最后通过4个图像数据集(MNIST、FashionMNIST、SVHN和CIFAR10)进行了实验，实验结果表明，相对于其他方法，本文所提出的方法在只有少量标签样本的情况下，可以获得更高的分类精度。 Finally, through a large number of experiments on four image datasets(MNIST, Fashion-MNIST, SVHN, and CIFAR10), the results show that ICP-ASSDL can achieve higher classification accuracy than other methods with a small number of labeled samples.
53003 传统的虚拟路径漫游是在固定方向路径上的漫游，而实时任意方向上的虚拟路径漫游在路径曲线处有抖动问题且计算量大，使得虚拟现实(V R)场景运行缓慢、晃动和卡顿，给人带来视觉上的晕眩感。 The traditional virtual path roaming is to roam on the fixed direction path, while the real-time virtual path roaming in any direction has jitter problem at the path curve, and the amount of calculation is large, which makes the virtual reality(VR) scene running slowly, shaking and stuck, and brings visual dizziness.
53004 针对以上问题提出了基于Dynapath算法、微分路径及改进的Cardinal曲线插值算法相结合的优化算法。 Aiming at the above problems, a new interpolation optimization algorithm is proposed based on Dynapath algorithm, differential path and improved Cardinal curve interpolation algorithm.
53005 使用Dynapath算法规划实时路径和等时间微分路径，并对微分线段等截距处进行初步平滑处理。 The Dynapath algorithm is used to plan the real-time path and the equal time differential path, and the differential line segment with equal intercept is smoothed preliminarily.
53006 最后对该路径使用改进的Cardinal曲线插值算法拟合出该曲线。 Finally, the improved Cardinal curve interpolation algorithm is used to fit the path.