Description: 提出了一种基于样本的分级检索
MPEG 视频的新方法:首先用I 帧的dct_dc_size 字段快速粗检,然后用断层摄影(tomography)法分析B 帧运动矢
量的时空分布特性以进一步缩小结果集,最后用DC 图像的精确匹配方法验证检索结果.试验结果表明,本方法
所需计算量较小,且可保证较高的检索精度.-presents a sample classification based on MPEG video retrieval of the new method : First of all I frame with the rapid field dct_dc_size rough Frederick, and then use tomography (tomography) Analysis of B pictures motion vector space-time distribution to further narrow the results set, and DC Images precision Matching test results retrieval. test results showed that the calculation method for small and can guarantee higher retrieval accuracy. Platform: |
Size: 501988 |
Author:隼鼻梁 |
Hits:
Description: 人脸检测源代码. The souce demonstrates face detection SSE optimized C++ library for color and gray scale data with skin detection, motion estimation for faster processing, small sized SVM and NN rough face prefiltering, PCA/LDA/ICA/any dimensionality reduction/projection and final NN classification Platform: |
Size: 366639 |
Author:huangtyu |
Hits:
Description: 提出了一种基于样本的分级检索
MPEG 视频的新方法:首先用I 帧的dct_dc_size 字段快速粗检,然后用断层摄影(tomography)法分析B 帧运动矢
量的时空分布特性以进一步缩小结果集,最后用DC 图像的精确匹配方法验证检索结果.试验结果表明,本方法
所需计算量较小,且可保证较高的检索精度.-presents a sample classification based on MPEG video retrieval of the new method : First of all I frame with the rapid field dct_dc_size rough Frederick, and then use tomography (tomography) Analysis of B pictures motion vector space-time distribution to further narrow the results set, and DC Images precision Matching test results retrieval. test results showed that the calculation method for small and can guarantee higher retrieval accuracy. Platform: |
Size: 501760 |
Author:隼鼻梁 |
Hits:
Description: 为了更准确地识别人的表情，在识别人脸7 种基本表情（愤怒、厌恶、恐惧、高兴、无表情、悲伤和惊讶）时，采用了局域二值
模式技术提取面部特征，进行由粗略到精细的表情分类。在粗略分类阶段，7 种基本表情中的2 种表情被选为初步分类结果（候选表情）。
在精细分类阶段，选用计算加权卡方值确定最终分类结果。采用日本的Jaffe 表情数据库来验证算法性能，对陌生人表情的识别率为77.9％，
其结果优于采用同样数据库的其他方法，且易于实现-In order to more accurately identify the person s facial expressions, in the identification of seven kinds of basic facial expressions (anger, disgust, fear, happy, expressionless, sadness and surprise), the use of local binary pattern of facial features extraction techniques, carried out by the rough to the fine expression classification. In the rough classification stage, the seven kinds of basic expressions in the two kinds of expression was selected as the initial classification results (candidate expressions). In the fine classification stage, the choice of calculating the weighted chi-square value to determine the final classification results. Jaffe expressions used in Japan to validate algorithm performance database of strangers face recognition rate was 77.9, the result is better than using the same database in other ways, and are easy to achieve Platform: |
Size: 212992 |
Author:张波 |
Hits:
Description: 文本分类综述和一些近期基于粗糙集的文本分类文章，其中一篇国防科技大学的关于文本分类的文章非常好，另外还有一篇清华大学博士毕业论文（关于机器学习文本分类）对毕业设计也是很有指导作用的。-Summary of text classification and some recent Rough set-based text classification article, of which a National University of Defense Technology of Text Classification article very good, as well as a doctoral thesis, Tsinghua University (on machine learning text classification) on the graduation project is also very guiding role. Platform: |
Size: 26036224 |
Author:willee |
Hits:
Description: 人脸检测源代码. The souce demonstrates face detection SSE optimized C++ library for color and gray scale data with skin detection, motion estimation for faster processing, small sized SVM and NN rough face prefiltering, PCA/LDA/ICA/any dimensionality reduction/projection and final NN classification -Face detection source code. The souce demonstrates face detection SSE optimized C++ Library for color and gray scale data with skin detection, motion estimation for faster processing, small sized SVM and NN rough face prefiltering, PCA/LDA/ICA/any dimensionality reduction/projection and final NN classification Platform: |
Size: 366592 |
Author:huangtyu |
Hits:
Description: 对文件进行分类，可以进行拖放，支持文件夹选择，开发工具是VS2008，拖放对于快捷方式无发得到目标路径，界面较粗糙，对于双击打开支持不够完全-Classification of documents, you can drag and drop, support for folder selection, development tools, is VS2008, drag and drop the shortcut target path of non-fat interface than the rough, not enough support for double-click to open fully Platform: |
Size: 4726784 |
Author:小鹏 |
Hits:
Description: 粗糙集是一种数据预处理算法，可以用遗传算法对其进行优化。是很好的分类算法-Rough set is a data pre-processing algorithm, genetic algorithm can be used for its optimization. Is a good classification algorithm Platform: |
Size: 12288 |
Author:王淑娟 |
Hits:
Description: 在基于特征提取方法之上, 研究用算法对目标多视角问题进行分类器设计。在对
图像进行独立成分分析后, 针对多姿态角目标识别问题, 提出了角度优先粗分类的设计方法-Feature extraction method based on above research objective of the algorithm to classify multi-view design problem. In the right image independent component analysis, the attitude angle for multi-target recognition problem, a point of first rough classification of design method Platform: |
Size: 720896 |
Author:f0700 |
Hits:
Description: 基于、的特征提取方法, 研究、算法的多角度分类器设计问
题在对图像进行主成分分析和独立成分分析后, 针对多姿态角目标识别问题, 提出了角度优先粗分
类的设计方法, 并给出系统流程, 最后-Based, feature extraction methods, research, multi-angle algorithms classifier design problem in the image principal component analysis and independent component analysis, the attitude angle for multi-target recognition problem, a point of first rough classification of design methods, and to give out of the system processes, the final Platform: |
Size: 323584 |
Author:f0700 |
Hits:
Description: 图像场景分类中视觉词包分类的应用与操作代码-Review of the bag-of-visual-words models in image scene classification
Platform: |
Size: 1112064 |
Author:libin |
Hits:
Description: 基于粗糙集的分类规则提取和分类规则约简。首先进行粗糙分类器训练，然后将复杂的分类规则降维成简单的分类规则，但分类精度不变。-Rough classification learning rules are trained and extracted based on rough sets and complex classification rules are dimensionally reduced into simple classification rules, but the classification accuracy are the same.
Platform: |
Size: 2048 |
Author:王长忠 |
Hits:
Description: 这是天津大学胡清华老师在粗糙集邻域领域做的最经典的源码，同学们可以在此基础上学习和修改，入口程序已经写好，需要其他方法可以自己添加，MAIN.m是入口程序，参数的意思在子函数里讲的很明白，调用了featureselect_FW_fast.m用来属性约简，几个clsf_dpd文件是使用不同的距离公式来计算属性重要度，选择得到属性结果，使用crossvalidate.m十折交叉算法来计算计算算法精度，该段代码调用了几个分类器，C4_5.m是决策树，KNN.m是最近邻分类器，NEC.m是类似于KNN的胡修改的程序，osu_svm3.00文件夹是使用svm分类器调用的文件，使用该分类器时，代码中间的路径需要修改。另外附上一堆常用的数据集。-This is Hu Qinghua teacher at Tianjin University neighborhood rough set field do the most classic source code, students can learn and modify On this basis, the entry procedures have been written, you need other ways to add your own, MAIN.m entry program is meaning parameters Functions talked in very clear call for the featureselect_FW_fast.m attribute reduction, several clsf_dpd file is to use a different formula to calculate the distance attribute importance, choose properties to get the results, use crossvalidate.m ten fold cross algorithm to calculate the accuracy of the calculation algorithm, the segment code calls several classifiers, C4_5.m is a decision tree, KNN.m is the nearest neighbor classifier, NEC.m is similar to KNN Hu modified program, osu_svm3. 00 folders using svm classifier called file, using the classification code in the middle of the path need to be modified. Also attach a bunch of common data sets. Platform: |
Size: 2542592 |
Author:robert |
Hits:
Search - rough classification