资源简介
发表在IJCV上选择性搜索算法的matlab代码,亲测可用。
代码片段和文件信息
function [abo mabo boScores avgNumSegments] = BlobAverageBestOverlap(gtBlobs gtNrs blobIndIm blobBoxes hierarchy minWidth)
% [abo mabo boScores avgNumSegments] = BlobAverageBestOverlap(gtBlobs
% gtNrs blobIndIm blobBoxes hierarchy minWidth)
%
% Calculate Average Best Overlap scores
%
% gtBlobs: Cell array of ground truth segments per class (see
% GetAllobjectBoxes)
% gtNrs: Cell array with image nrs corresponding to ground truth.
% blobIndIm: Image with indices per blob (mexFelzenSegmentIndex)
% blobBoxes: Boxes corresponding to blobs in blobIndIm
% hierarchy: Hierarchy necessary to reconstruct all blobs in grouping
% minWidth: (optional) Filter out blobs with a width smaller than minWidth.
%
% abo: Average Best Overlap per class (Pascal Overlap criterion)
% mabo: Mean Average Best Overlap (mean(abo))
% boScores: Best Overlap Score per GT segment.
% avgNumBlobs: Average number of blobs per image
%
% Jasper Uijlings - 2013
if ~exist(‘minWidth‘ ‘var‘)
minWidth = 0;
end
nClasses = length(gtBlobs);
% Memory initialization
numSegments = zeros(length(blobIndIm) 1);
boScores = cell(1 nClasses);
for cI = 1:nClasses
boScores{cI} = length(gtBlobs{cI});
end
% indices per class
classIdx = ones(1 nClasses);
for cI=1:length(gtBlobs)
for i=1:length(gtBlobs{cI})
testImNr = gtNrs{cI}(i);
% the hierarchy here contains possibly multiple groupings with
% different initial measures
testBlobsT = cell(length(hierarchy{testImNr}) 1);
testBlobsT{1} = RecreateBlobHierarchyIndIm(blobIndIm{testImNr} blobBoxes{testImNr} hierarchy{testImNr}{1});
for j=2:length(hierarchy{testImNr}) % Without initial blobs here
[aa bb testBlobsT{j}] = RecreateBlobHierarchyIndIm(blobIndIm{testImNr} blobBoxes{testImNr} hierarchy{testImNr}{j});
end
testBlobs = cat(1 testBlobsT{:});
% Get rid of too small blobs
testBlobs = FilterBlobsWidth(testBlobs minWidth);
numSegments(testImNr) = length(testBlobs);
% Calculate overlap scores
boScores{cI}(classIdx(cI)) = BlobBestOverlap(testBlobs gtBlobs{cI}(i));
classIdx(cI) = classIdx(cI) + 1;
end
end
abo = zeros(nClasses 1);
for cI = 1:nClasses
abo(cI) = mean(boScores{cI});
end
mabo = mean(abo);
% Average of numSegments. Make sure that only images for which the
% numSegments are actually calculated are taken into account.
avgNumSegments = mean(numSegments(numSegments > 0));
属性 大小 日期 时间 名称
----------- --------- ---------- ----- ----
....... 90230 2007-01-09 18:38 SelectiveSearchCodeIJCV\000015.jpg
文件 10758 2016-10-23 23:59 SelectiveSearchCodeIJCV\airplane_056.jpg
文件 19521 2016-10-24 12:56 SelectiveSearchCodeIJCV\airport_101.jpg
....... 15510 2013-05-31 23:45 SelectiveSearchCodeIJCV\anigauss.mexa64
文件 25600 2017-07-02 16:17 SelectiveSearchCodeIJCV\anigauss.mexw64
....... 2611 2013-06-14 22:06 SelectiveSearchCodeIJCV\BlobAverageBestOverlap.m
....... 751 2013-06-14 22:06 SelectiveSearchCodeIJCV\BlobBestOverlap.m
....... 922 2013-05-31 23:33 SelectiveSearchCodeIJCV\BlobStruct2HierarchicalGrouping.p
....... 316 2013-05-31 23:33 SelectiveSearchCodeIJCV\BlobStructColourHist.p
....... 424 2013-05-31 23:33 SelectiveSearchCodeIJCV\BlobStructTextureHist.p
....... 1447 2013-06-14 22:06 SelectiveSearchCodeIJCV\BoxAverageBestOverlap.m
....... 187 2013-05-31 23:33 SelectiveSearchCodeIJCV\ChangeEdges.p
文件 19850 2016-10-24 13:04 SelectiveSearchCodeIJCV\chaparral_055.jpg
文件 11647 2016-10-24 00:05 SelectiveSearchCodeIJCV\cloud_109.jpg
文件 2953 2017-07-03 19:06 SelectiveSearchCodeIJCV\demo.m
....... 4346 2013-06-03 22:35 SelectiveSearchCodeIJCV\demoPascal2007.m
....... 19447 2013-05-31 23:33 SelectiveSearchCodeIJCV\Dependencies\anigaussm\anigauss.c
....... 4481 2013-05-31 23:33 SelectiveSearchCodeIJCV\Dependencies\anigaussm\anigauss.m
....... 1753 2013-05-31 23:33 SelectiveSearchCodeIJCV\Dependencies\anigaussm\anigauss_mex.c
....... 203 2013-05-31 23:33 SelectiveSearchCodeIJCV\Dependencies\Blob2Image.p
....... 200 2013-05-31 23:33 SelectiveSearchCodeIJCV\Dependencies\Blob2Vector.p
....... 138 2013-05-31 23:33 SelectiveSearchCodeIJCV\Dependencies\BlobAddSizes.p
....... 220 2013-05-31 23:33 SelectiveSearchCodeIJCV\Dependencies\BlobAddTextureHists.p
....... 148 2013-05-31 23:33 SelectiveSearchCodeIJCV\Dependencies\Blobs2Boxes.p
....... 713 2013-06-14 22:07 SelectiveSearchCodeIJCV\Dependencies\BoxBestOverlap.m
....... 645 2013-06-14 22:07 SelectiveSearchCodeIJCV\Dependencies\BoxIntersection.m
....... 536 2013-06-14 22:07 SelectiveSearchCodeIJCV\Dependencies\BoxRemoveDuplicates.m
....... 625 2013-06-14 22:07 SelectiveSearchCodeIJCV\Dependencies\BoxSize.m
....... 371 2013-06-14 22:07 SelectiveSearchCodeIJCV\Dependencies\BoxUnion.m
....... 1108 2013-06-14 22:07 SelectiveSearchCodeIJCV\Dependencies\CountVisualWordsIndex.m
............此处省略65个文件信息
- 上一篇:matlab各个版本安装包以及破解文件
- 下一篇:MATLAB图片批量归一化
相关资源
- 机器学习好瓜坏瓜分类
- [精品]matlab仿真A*路径规划算法,100
- matlab与机器学习百度云分享
- MATLAB与机器学习详细.txt
- 用matlab从0开始实现的一个多层感知机
- adaboost 演示demo基于Matlab,学习算法包
- logistic回归matlab
- 随机森林matlab代码分类RF/回归RF
- 支持向量机SVM机器学习方法
- 机器学习、人工智能、数据挖掘中经
- 马尔科夫链蒙特卡洛MCMC仿真带MATLAB代
- 增广拉格朗日乘子法ALM算法matlab代码
- MATLAB 2016b 安装包
- RBF and svm matlab code matlab回归预测的源
- BP神经网络进行多分类matlab代码 (c
- Fisher-Score机器学习
- RVM2 基于稀疏贝叶斯框架的机器学习算
- MATLAB 神经网络43个案例分析
- popular-UCI-datasets 一些非常有用的数据
- Machine-Learning-exercises_finished andrew NG上
- Machine-Learning 《机器学习》
- spider 机器学习matlab源代码
- psoSVMcgForClass.m
- HOG+SVM图像分类算法
- 遗传算法求解0-1背包问题matlab代码.
- BP神经网络matlab程序
- 基于形态学的权重自适应图像去噪
- 粒子群算法的彩色图像分割(聚类)
- matlab实现决策
- 单类支持向量机
评论
共有 条评论