;; UPDATE: For KITTI users, you may find `br-dired-kitti-label-to-xml' useful

;; Usage for KITTI dataset (M-x means Alt+x):

;; (1) open this file in emacs.

;; (2) modify the above (setq kitti-image-root "/home/user/path-to-kitti-root/training/image/");

;; to your path to KITTI image root

;; (3) Run M-x eval-buffer or (load-file "path-to-your/table-xml.el") to load all the functions in

;; this file as well as the kitti-image-root path (after you have changed it to the correct path).

;; (4) M-x dired (go to label folder of KITTI)

;; (5) `dired-mark' or `dired-mark-files-regexp' function to mark the label files that you want to convert to XML

;; Note: do not mark too many at a time. About 2000 files a time would be good. I experienced some crashes

;; and memory problems with too many files at a time. It's probably due to an emacs internal problem.

;; (6) M-x `br-dired-kitti-label-to-xml'

;; (7) xml files will be created in ./xml folder

;; (optional) change function `etxml-KITTI-print-buffer-to-xml' and etxml-KITTI-dataset-schema to fit

;; your need. Read the below advanced tutorial if you want to know how it works. If you

;; want to customize the output XML file to some other formats, modify etxml-KITTI-dataset-schema

;; variable below.

 

1 首先在ubuntu下面安装编辑器emacs ：使用命令 sudo apt-get install emacs24

2 运行命令emacs打开程序，打开 exXML_0.47 ，修改KITTI的路径

3 按下ALT+X，然后输入 eval-buffer 加载文件中的函数。

4 按下ALT+X ，输入dired，进入KITTI的label文件夹

5 按下ALT+X 输入dired-mark或者用   “dired-mark-files-regexp” 正则化匹配文件     （怎么选前2000个还不会）

6 按下ALT+X ，输入br-dired-kitti-label-to-xml，运行就转换成功。

 

 

这个网址   有使用python转换的教程。

 

 

二、 PASCAL VOC 11point 方法

1 首先对所有预测结果进行排序，score分高的排在前面，分数低的排在后面

2 按顺序逐个把样本作为正例进行预测，计算出FP和TP

3 再计算 召回率rec和精度prec

4 按照召回率t从0到1之间每0.1一个点，一共11个点，分别计算大于t的最大精度值p ，然后按照 ap=ap+p/11 ，一共计算11次。

5 ap即为检测的值，所有的ap平均值为mAP。