【ACF学习】ACF安装、训练及检测初探

使用环境说明：Win10+VS2013+MATLAB R2016a

1，dollar工具箱下载编译

https://pdollar.github.io/toolbox/，最新版本是Version3.5

For version history click here. This code is licensed under the Simplified BSD License.

• Recommended: download the latest version from GitHub
• Version 3.50   May 12, 2016
• Version 3.00   Aug. 06, 2012
• Version 2.00   Sep. 30, 2007
• Version 1.03   May. 03, 2006   [compatible with cuboids code]
• Version 1.00   Oct. 02, 2005

下载后添加目录、编译即可使用

addpath(genpath('ROOT\piotr_toolbox\toolbox\'))

savepath

toolboxCompile

注意编译前需要配置好VS编译器

工具箱包含七大部分，channels、classify、detector、filter、images、matlab、videos，后续会对每部分做详细的说明介绍

2，测试已训练模型

可以准备好Inria或者Caltech数据使用detector中acfDemoInria.m或者acfDemoCal.m测试，也可以直接用图片测试

load('detector/models/AcfInriaDetector.mat')

加载的detector分为三部分：opts、clf、info

opts保存模型相关信息

clf保存分类器信息

info保存特征通道信息

image = imread('test.jpg')

box = acfDetect(image, detector)

figure;

im(image);

bbApply('draw', box);

% The detector was introduced and described through the following papers:
%  [1] P. Doll醨, Z. Tu, P. Perona and S. Belongie
%   "Integral Channel Features", BMVC 2009.
%  [2] P. Doll醨, S. Belongie and P. Perona
%   "The Fastest Pedestrian Detector in the West," BMVC 2010.
%  [3] P. Doll醨, R. Appel and W. Kienzle
%   "Crosstalk Cascades for Frame-Rate Pedestrian Detection," ECCV 2012.
%  [4] P. Doll醨, R. Appel, S. Belongie and P. Perona
%   "Fast Feature Pyramids for Object Detection," PAMI 2014.
%  [5] W. Nam, P. Doll醨, and J.H. Han
%   "Local Decorrelation For Improved Pedestrian Detection," NIPS 2014.
% Please see: http://vision.ucsd.edu/~pdollar/research.html#ObjectDetection