搜索资源列表
jiyufenkuaizhifangtuhejiaohuxinxidejintoujiance.ra
- 本文提出了一种新的基于对模型的分块直方图求交互信息量的镜头检测算法。算法中对每帧图象 进行分块求直方图, 然后利用相邻帧间对应分块的直方图统计值求交互信息量, 最后把所有分块的交互信息量进行 加权平均以检测镜头的变化。实验结果表明与传统的直方图相比, 该算法对一般场景的镜头变化有更高的查全率和 查准率。-This paper presents a new model based on the sub-block histogram information for cross-dete
Text-Retrieval
- 信息检索系统从最初的纯手工检索系统业已发展到现在的以信息技术为支撑的检索系统,在这一过程中,适应新的信息资源、信息技术这些检索环境,提高信息检索系统的查全率、查准率和系统响应时间是不变的主题,在众多文本中掌握最有效的信息始终是信息处理的一大目标。围绕向量空间模型设计了一个文本检索系统,介绍向量空间模型的基础上给出了基于它的信息检索系统的一般结构框架和各部分的功能,探讨了系统中所涉及到的关键技术。用向量空间模型进行特征表达,用TF-IDF(Term-Frequency Inverse-Docume
A-SHOT-BOUNDARY-DETECTION-ALGORITHM-BASED-ON
- Shot boundary detection is always an important topic in digital video processing. It is the first important task of content-based video retrieval and indexing. In this paper, a new shot boundary detection algorithm is proposed, based on Parti
New-folder
- how to normalize a data, calculation of local measure, precision recall formula,-how to normalize a data, calculation of local measure, precision recall formula,
New-folder1
- local descr iptor, normalization of date, precision recall calculation
labinithom
- We will use MatCont to start homoclinic orbits that emanate BT points in several models 1 . Recall that in the GUI of MatCont k is denoted as TTolerance. In all examples, we will set TTolerance=10 − 5 (k). As a rule, the amplitude (A) shou
PE_doc_29919
- Commonly used uation measures including Recall, Precision, F-Measure and Rand Accuracy are biased and should not be used without clear understanding of the biases, and corresponding identification of chance or base case levels of the statistic.
Robust_Face_Landmark
- 在现实世界条件下获取人脸存在较大的变化在形状和遮挡由于不同在姿态、表情、附属品的使用,例如,太阳镜和帽子以及与目标体(e.g. 食物)的交。当前的人脸界标估计方法在这种条件下努力但由于缺乏一种有效的理论方法用于处理局外点。我们提供了一个新奇的方法,称为Robust Cascaded Pose Regression (RCPR),通过检测显式的遮挡且使用鲁棒的形状索引的特征可以减少exposure对于局外点。我们证明RCPR改进先前的界标估计方法在3个通用的人脸数据集上(LFPW, LFW and