人脸检测——FDDB数据集评估

人脸检测——FDDB数据集评估

一、FDDB数据集简介

FDDB数据集官网：http://vis-www.cs.umass.edu/fddb/index.html

• FDDB数据集用于人脸检测算法。This data set contains the annotations for 5171 faces in a set of 2845 images taken from the Faces in the Wild data set。他的人脸原始标签的box是椭圆型，所以评估时候有rectangle和椭圆两种方式。
• 图片文件夹：包含两个文件夹2002和2003两个文件夹
• ├── 2002
  │ ├── 07
  │ ├── 08
  │ ├── 09
  │ ├── 10
  │ ├── 11
  │ └── 12
  ├── 2003
  │ ├── 01
  │ ├── 02
  │ ├── 03
  │ ├── 04
  │ ├── 05
  │ ├── 06
  │ ├── 07
  │ ├── 08
  │ └── 09
• 标签文件 FDDB-folds文件夹，包括20个txt文件。
• FDDB-fold-01-ellipseList.txt 内容格式：

FDDB-fold-xx-ellipseList.txt" in the following 
format:
...
<image name i>
<number of faces in this image =im>
<face i1>
<face i2>
...
<face im>
...

number of faces in this image =im 代表这张图的人脸个数， each face 是:
<major_axis_radius minor_axis_radius angle center_x center_y 1>.
例如：
2002/08/11/big/img_591
1
123.583300 85.549500 1.265839 269.693400 161.781200 1

• 另外一种txt是只包含图片路径：比如：2003/01/15/big/img_763
  FDDB-folds
  │ ├── FDDB-fold-01-ellipseList.txt
  │ ├── FDDB-fold-01.txt
  │ ├── FDDB-fold-02-ellipseList.txt
  │ ├── FDDB-fold-02.txt
  │ ├── FDDB-fold-03-ellipseList.txt
  │ ├── FDDB-fold-03.txt
  │ ├── FDDB-fold-04-ellipseList.txt
  │ ├── FDDB-fold-04.txt
  │ ├── FDDB-fold-05-ellipseList.txt
  │ ├── FDDB-fold-05.txt
  │ ├── FDDB-fold-06-ellipseList.txt
  │ ├── FDDB-fold-06.txt
  │ ├── FDDB-fold-07-ellipseList.txt
  │ ├── FDDB-fold-07.txt
  │ ├── FDDB-fold-08-ellipseList.txt
  │ ├── FDDB-fold-08.txt
  │ ├── FDDB-fold-09-ellipseList.txt
  │ ├── FDDB-fold-09.txt
  │ ├── FDDB-fold-10-ellipseList.txt
  │ ├── FDDB-fold-10.txt

二、评估流程

• 1、下载数据和评估代码，数据就是上面图片文件夹和标签文件夹

• 2、下载评估代码文件：evaluation

• 3、编译evaluation代码（ubuntu16.04下）

  • 系统环境：opencv3.2.0 gnuplot
    安装gnuplot

  sudo apt-get install gnuplot
  

  • 编译evaluation
    1）修改Makefile文件

    # 1、加入下面内容
    INCS = -I/usr/local/include/opencv
    
    LIBS = -L/usr/local/lib -lopencv_core -lopencv_imgproc -lopencv_highgui
           -lopencv_ml -lopencv_video -lopencv_features2d -lopencv_calib3d 
           -lopencv_objdetect -lopencv_contrib -lopencv_legacy
    		
    # 2、修改 evaluate.， $(CC) $(OBJS) -o $@ $(LIBS)这一句前面Tab键自己输入下，也就是空格，要不然可能报错：***missing separator. Stop
    
    evaluate: $(OBJS)
    
        	$(CC) $(OBJS) -o $@ $(LIBS)
    

    2）修改common.hpp 13行，ppm改为jpg

    #define __IMAGE_FORMAT__ ".jpg"
    //#define __IMAGE_FORMAT__ ".ppm"
    #define __CVLOADIMAGE_WORKING__
    

    3）在evaluation 目录下

    make
    

    • 自己算输出文件的个格式：有十个文件，命名为下面格式，分别对用上面标签文件的图片内容。
      ├── fold-01-out.txt
      ├── fold-02-out.txt
      ├── fold-03-out.txt
      ├── fold-04-out.txt
      ├── fold-05-out.txt
      ├── fold-06-out.txt
      ├── fold-07-out.txt
      ├── fold-08-out.txt
      ├── fold-09-out.txt
      └── fold-10-out.txt
      文件内容格式：

			...
			<image name i>
			<number of faces in this image =im>
			<face i1>
			<face i2>
			...
			<face im>
			...

      face im格式，有两种：

	      4 a. Rectangular regions，方框
		       Each face region is represented as:
		       <left_x top_y width height detection_score> 
		  
		  4 b. Elliptical regions，椭圆
		       Each face region is represented as:
		       <major_axis_radius minor_axis_radius angle center_x center_y detection_score>.

• 例如：
  2002/08/11/big/img_591
  1
  192.178558 60.700516 150.671692 210.003891 0.998205
  2002/08/26/big/img_265
  3
  302.738037 58.903488 79.118591 124.53154 0.999873
  63.425415 33.952583 84.562866 119.966179 0.999395
  162.020691 101.475128 45.098541 75.268555 0.994589

• 3 开始评估

  • 1）修改runEvaluate.pl文件

  my $GNUPLOT = "/usr/bin/gnuplot"; # 修改为gnuplot库目录
  # where the binary is
  my $evaluateBin = "./evaluate"; 
  # where the images are
  my $imDir = "../"; # 图片目录，2002文件夹上一级目录
  # where the folds are
  my $fddbDir = "../FDDB-folds/"; # 标签目录
  # where the detections are
  my $detDir = "../resulst/" # fold-01-out.txt所在文件夹的目录
  

  • 2 )运行
    perl runEvaluate.pl

• 生成四个文件

  ContROC.png
  ContROC.txt*
  DiscROC.png
  DiscROC.txt
  

• roc curve DiscROC.png*