在windows平台上测试自己的人脸检测算法在FDDB数据集

弄了好几天终于能在FDDB数据集上测试自己用wider face数据集训练的的faster rcnn的检测器的性能了。

1、首先在官网http://vis-www.cs.umass.edu/fddb/index.html下载图片和标注文件，FDDB的标注文件，分为图片名称文件FDDB-fold-01.txt、对应的标注文件：FDDB-fold-01-ellispe.txt，各有10个，把10个文件按顺序合并，分别命名为Fold_all.txt和Elsp.txt.

2、读Fold.txt文件，依照顺序做人脸检测，将检测结果输出出来（
4000
我是将faster rcnn中test.py文件修改得到的检测结果）。

3、我的mxnet的faster rcnn生成的检测结果格式为<image name i> score x1 y1 x2 y2,与FDDB要求的格式http://vis-www.cs.umass.edu/fddb/README.txt不一致，所以写了个C++程序将我的输出格式转化为FDDB要求的格式

// ToFDDBResults.cpp : Defines the entry point for the console application.
//

#include "stdafx.h"
#include<iostream>
#include<fstream>
#include<sstream>
#include<string>
#include<vector>
using namespace std;

int _tmain(int argc, _TCHAR* argv[])
{
ifstream InputFile("comp4_det_Foldall_face.txt");
ofstream OutputFile("comp4_det_Foldall_face_results.txt");
string line, ImageNameTemp;
stringstream ss;
int count = 1;
vector<double> vec_score;
vector<double> vec_x1;
vector<double> vec_y1;
vector<double> vec_x2;
vector<double> vec_y2;

double score, x1, y1, x2, y2;
string ImageName;
getline(InputFile, line);
ss << line;
ss >> ImageName >> score >> x1 >> y1 >> x2 >> y2;
vec_score.push_back(score);
vec_x1.push_back(x1);
vec_y1.push_back(y1);
vec_x2.push_back(x2);
vec_y2.push_back(y2);
ImageNameTemp = ImageName;

while (getline(InputFile, line))
{
ss.str("");
ss.clear();
ss << line;
ss >> ImageName >> score >> x1 >> y1 >> x2 >> y2;
if (ImageName == ImageNameTemp)
{
count++;
vec_score.push_back(score);
vec_x1.push_back(x1);
vec_y1.push_back(y1);
vec_x2.push_back(x2);
vec_y2.push_back(y2);
}
else
{
OutputFile << ImageNameTemp << endl;
OutputFile << count << endl;
count = 1;
for (int i = 0; i < vec_score.size(); i++)
{
OutputFile << vec_x1[i] << " " << vec_y1[i] << " " << (vec_x2[i] - vec_x1[i]) << " "
<< (vec_y2[i] - vec_y1[i]) << " " << vec_score[i] << endl;

}
vec_score.clear();
vector<double>(vec_score).swap(vec_score);
vec_x1.clear();
vector<double>(vec_x1).swap(vec_x1);
vec_y1.clear();
vector<double>(vec_y1).swap(vec_y1);
vec_x2.clear();
vector<double>(vec_x2).swap(vec_x2);
vec_y2.clear();
vector<double>(vec_y2).swap(vec_y2);
vec_score.push_back(score);
vec_x1.push_back(x1);
vec_y1.push_back(y1);
vec_x2.push_back(x2);
vec_y2.push_back(y2);
}
ImageNameTemp = ImageName;
}
OutputFile << ImageName << endl;
OutputFile << count << endl;
for (int i = 0; i < vec_score.size(); i++)
{
OutputFile << vec_x1[i] << " " << vec_y1[i] << " " << (vec_x2[i] - vec_x1[i]) << " "
<< (vec_y2[i] - vec_y1[i]) << " " << vec_score[i] << endl;

}

return 0;
}

4、在http://vis-www.cs.umass.edu/fddb/results.html下载fddb evaluate代码，新建一个vs工程，配置好OPENCV，把evaluate中的源文件和头文件加进来，修改evaluate.cpp文件73行

string baseDir = "C:/Users/qingyun.tang/Documents/Visual Studio 2013/Projects/FDDB_testing/FDDB_testing/";//FDDB图片路径，此路径下需包含文件夹2002，2003
string listFile = "C:/Users/qingyun.tang/Documents/Visual Studio 2013/Projects/FDDB_testing/FDDB_testing/fddb/Fold_all.txt";
string detFile = "C:/Users/qingyun.tang/Documents/Visual Studio 2013/Projects/FDDB_testing/FDDB_testing/fddb/comp4_det_Foldall_face_results.txt";//转化了格式后的结果文件
string annotFile = "C:/Users/qingyun.tang/Documents/Visual Studio 2013/Projects/FDDB_testing/FDDB_testing/fddb/Elsp.txt";

因为我直接在vs上运行，不在cmd上运行这个evaluate程序，所以注释以下几行 代码：

//if(argc == 1)
//{
//  printUsage();
//  return 0;
//}

在项目的属性栏中做如下改动

编译，通过。生成ContROC.txt和 DiscROC.txt两个文件

5、安装GUNPLOT，我装的gp501-win32-mingw，

配置perl，参考https://jingyan.baidu.com/article/9f7e7ec0b798ae6f281554e9.html?st=2&os=0&bd_page_type=1&net_type=1安装好之后，修改下载的evaluate代码中的runEvaluate.pl文件：

#!/usr/bin/perl -w

use strict;

#### VARIABLES TO EDIT ####
# where gnuplot is
my $GNUPLOT = "C:/Program Files (x86)/gnuplot/bin/gnuplot"; #gnuplot安装路径
# where the binary is
my $evaluateBin = "evaluate";
# where the images are
my $imDir = "C:/Users/qingyun.tang/Documents/Visual Studio 2013/Projects/FDDB_testing/FDDB_testing"; #[code]FDDB图片路径

# where the folds are
my $fddbDir = "C:/Users/qingyun.tang/Documents/Visual Studio 2013/Projects/FDDB_testing/FDDB_testing/FDDB-folds.txt";
# where the detections are
my $detDir = "C:/Users/qingyun.tang/Documents/Visual Studio 2013/Projects/FDDB_testing/FDDB_testing/fddb/";
###########################

my $detFormat = 0; # 0: rectangle, 1: ellipse 2: pixels

sub makeGNUplotFile
{
my $rocFile = shift;
my $gnuplotFile = shift;
my $title = shift;
my $pngFile = shift;

open(GF, ">$gnuplotFile") or die "Can not open $gnuplotFile for writing\n";
#print GF "$GNUPLOT\n";
print GF "set term png\n";
print GF "set size 1,1\n";
print GF "set output \"$pngFile\"\n";
#print GF "set xtics 500\n";
print GF "set ytics 0.1\n";
print GF "set grid\n";
#print GF "set size ratio -1\n";
print GF "set ylabel \"True positive rate\"\n";
print GF "set xlabel \"False positives\"\n";
#print GF "set xr [0:2000]\n";
print GF "set yr [0:1.0]\n";
print GF "set key right bottom\n";
print GF "plot \"$rocFile\" using 2:1 title \"$title\" with lines lw 2 \n";
close(GF);
}

my $gpFile = "C:/Users/qingyun.tang/Documents/Visual Studio 2013/Projects/FDDB_testing/FDDB_testing/fddb/ContROC.p";
my $gpFile1 = "C:/Users/qingyun.tang/Documents/Visual Studio 2013/Projects/FDDB_testing/FDDB_testing/fddb/DistROC.p";
my $title = "tqy";

# plot the two ROC curves using GNUplot
# plot the two ROC curves using GNUplot
makeGNUplotFile("C:/Users/qingyun.tang/Documents/Visual Studio 2013/Projects/FDDB_testing/FDDB_testing/tempContROC.txt", $gpFile, $title, $detDir."ContROC.png");
makeGNUplotFile("C:/Users/qingyun.tang/Documents/Visual Studio 2013/Projects/FDDB_testing/FDDB_testing/tempDiscROC.txt", $gpFile1, $title, $detDir."DiscROC.png");[/code]

在命令行中runEvaluate.pl的路径下输入perl runEvaluate.pl,得到ContROC.p与DistROC.p。

6、打开gnuplot软件：

a、file->open->contROC.p b、file->output->contROC.p

在对应目录中生成contROC.png;

a、file->open->discROC.p  b、file->output->discROC.p
在对应目录中生成discROC.png

7、如果想将自己的结果与官网上其他公司公布的结果画在一起，在官网上下载ContROC.p,DiscRoc.p等，用写字板打开之后，将自己的检测结果添加进去，同样用gnuplot画出来就行。