[OTB use introduction 2] Introduce the source code of Visual Tracker Benchmark v1.0

In [OTB Usage Introduction 1], I briefly explained how to run OTB. This time, I mainly introduce some settings of OTB and the resulting graphs in detail.

Among the quantitative evaluation indicators of OTB, there are three kinds of OPE, TRE and SRE, which can be obtained by perfPlot.m.
The visual bounding boxes of the video image are obtained by drawResultBB.m

The main function of main is to run the code of the tracking algorithm and store the results returned by the tracking algorithm in the result folder. The specific folder is determined by the following line of code. (The result is stored in the form of mat)

%运行main后得到的结果将存在此路径下（不修改就是在results文件夹的CVPR13下，可以自由修改)，
finalPath = ['./results/results_' evalType '_CVPR13/'];

evalType is to choose to run one of the three types of OPE, TRE, and SRE. As follows

%'OPE','SRE','TRE' 选择不同的评价标准， 初始为TRE 
%这里你想得到什么结果就修改为什么评价标准
evalType='TRE'; 

To explain here, OPE is a success rate, TRE (20 times) and SRE (12 times) are multiple times. The first result of TRE is actually the result of OPE, so you can run TRE and SRE once. That's about it for the changes in main.

After modifying main, run it and get the result. The results are shown below.
This is the basketball.mat file of CSK in results_TRE_CVPR13 in the results folder.

Click on the first one and you can see

these . The data in res is the result of the tracking algorithm . 725 is the frame number of the basketball sequence. The first line It is the first frame, and the 4 columns represent (x, y, w, h), where x and y represent the center of the target position, and w and h are the width and height of the target frame. When adding your own code to OTB later, you should pay special attention not to write w and h in reverse. Anno is the manually labeled information, that is, the reference data for the subsequent evaluation indicators.

Next is perfPlot.m, which is used to draw various curve graphs, that is, various success rate graphs and various accuracy graphs in the original paper. The pictures are saved in the figs folder.

evalTypeSet = {'SRE', 'TRE', 'OPE'};

This line of code controls several kinds of pictures, you can only keep the pictures you come up with, and delete the others.

rankingType = 'threshold';%AUC, threshold

It turned out to be AUC here, only the success rate (success) graph can be produced, if you want to come up with the precision (precision) graph, you need to modify it to threshold. This will give you both pictures.

Note that threshold is misspelled in the source comments. Correct it.

After modifying these, you can run perfPlot.m.

The last is drawResultBB.m, the result of the tracking algorithm in the obtained video is saved in the img of the tmp folder. As shown in the figure, the operation of

this m file requires the result of SRE, so before running, you need to select SRE and run the main file.
Then you can get the result graph like this.

pathRes = '.\results\results_SRE_CVPR17\';% The folder containing the tracking results

That is, in this folder, there are corresponding mat files.

[OTB usage introduction 2] is over here. [OTB usage introduction 3] introduces how to add your algorithm to OTB and compare it with other algorithms.