Series multi-target tracking algorithm: DMAN

Download Source:

https://github.com/jizhu1023/DMAN_MOT

Download the trained model, the new folder model, the model file in the model folder:

The authors used a python2.7, I use 3.5, it will print an error report, plus all brackets can be.

Then run the calculate_similarity.py in pycharm

In the current directory to the matlab DMAN_MOT-master:

Run DMAN_demo.m file, this time error: opt.mot does not exist, this variable is mot specify the path where the data set, so finding the definition of opt in, opt = globals (), open globals.m, add the following line:

Then re-run DMAN_demo.m file, this time will be given:

Cause of the error using mtimesx_build (line 166) AC / C ++ compiler has not been selected, it should be no designated compiler

Baidu, to find a solution:

https://blog.csdn.net/discoverer100/article/details/61195040?utm_source=blogxgwz4

I.e. matlab enter on command line: mex -setup C ++, to give the following output, indication vs2013 compiled.

Then continue to run DMAN_demo.m file, error:

Undefined function or variable 'MDP_value', this file is inside the MDP project, copied, but can not be used, parameters do not match, the message has been github author, and not reply. Ended!

---------------------------------------------------------------------------------------------------------------------------------------------------------------

Authors' reply has fixed the bug, I downloaded a new project, do a simple modification:

重新跑了一次,旧问题没了,新问题出现:

问题原因:是因为下面的判断

打印出flag,其内容是:b'client ok',前面多了个b,被编码成了byte类型,所以程序无法走到else的部分,而else下面的部分就是产生similarity.mat的,所以这里做修改:'client ok'前面加个b

接下来还会报byte和str的错,出错点如下所示

因为traj_dir编码错误,前面还是会有个b,所以无法按照str类型来切分,做下面修改即可解决问题:

接下来还要修改mod这个地方,因为python2和3的区别导致gap计算出来是float型的小数,而range中需要的是整数,所以将

改为双斜杠即可:

同理因为编码问题还需修改最后的地方:

网络编程使用的都是byte类型,所以这里要给前面加上b。

--------------------------------------------------------------------------------------------------------------------------------------------------------------

至此,所有错误都已经解决,可以愉快的输出结果了:

python端:

matlab端:

这里的从1到19的文件夹中保存的就是跟踪到的19个目标,每个文件夹中保存的是之前有跟踪到的帧中的该目标,示例如下:

 

Guess you like

Origin blog.csdn.net/sinat_33486980/article/details/95078922