Results are different with Python

[buingocnam87]

Hi! Thank you for providing such a good instruction of using faster rcnn with C++. I have successfully compiled the code but the results are really bad. Are there any issues? I used exactly your soft and change the caffe model only.

[Nvpiao]

I do have the same problem.

[zhanglaplace]

@buingocnam87 hi，i compiled the project failed.could you show me your experience for me ?

[Nvpiao]

@zhanglaplace
Here it‘s the blog writtend by author. I prefer u learn the first one，because the dynamic lib version may case much problem.
http://www.cnblogs.com/louyihang-loves-baiyan/p/5485955.html

[zhanglaplace]

@Lone5267 yes ,i have read the blog.because my operator system is window。
i attemp to use visual studio to read the python caffemodel , i build the caffe with python support
but it still have the error in create python layers.and i want to write a nms functions by myself。

[Nvpiao]

@zhanglaplace It may a tough project to transplant the code from Linux platform to window platform. I've been looking for solutions, but i have no idea about it. I just find the windows version of fast-rcnn.
I posted it here:
http://m.blog.csdn.net/article/details?id=52618386

may you find the way to solve faster-rcnn and i will fork you :)

[zhanglaplace]

thank you all the same. i want to have a try on windows.hope can success 2017-02-22 12:09 GMT+08:00 Lone5267 <[email protected]>:

…

@zhanglaplace <https://github.com/zhanglaplace> It may a tough project to transplant the code from Linux platform to window platform. I've been looking for solutions, but i have no idea about it. I just find the windows version of fast-rcnn. I posted it here: http://m.blog.csdn.net/article/details?id=52618386 may you find the way to solve faster-rcnn and i will fork you :) — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ALnWHlOoUmAqB4qSHLndIZ318lC3Q_Q2ks5re7T0gaJpZM4J3hYS> .

[handong1587]

I cannot get any correct detection results using this demo, it seems that it just produced some random bboxes. I was running this demo on Ubuntu 14.04.
Does anyone have a clue how to resolve this?

[zhanglaplace]

the platform is ubuntu or windows

…

---------------- Feng Zhang 张峰 School of Automation, Huazhong Univ. of Sci. & Tech. 1037 Luoyu Road, Wuhan 430074, P. R. China QQ:476833883 Mobile: +86-13129913417 2017-03-16 15:57 GMT+08:00 handong1587 <[email protected]>:

I cannot get any correct detection results using this demo, it seems that it just produced some random bboxes. Does anyone have a clue how to resolve this? [image: vis] <https://cloud.githubusercontent.com/assets/2013252/23986771/d8078bd2-0a60-11e7-9a67-1e49879d4ad8.jpg> — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ALnWHnW3NzODIzKLMh-MmXW94H3lWir3ks5rmOtMgaJpZM4J3hYS> .

[handong1587]

@zhanglaplace Hi, I was running this demo on Ubuntu 14.04.
Actually I once ported py-faster-rcnn to Windows, but got similar results. I am very confused why this happen :-(

[zhanglaplace]

@handong1587 我没有ubuntu环境，所以不清楚。但是我做了个windows VC++调用matlab模型的。

[handong1587]

@zhanglaplace
After made some modifications, I think I corrected the defects, now it can detect object correctly.
One of the modificaiton I made is the 'class_num' in faster_rcnn.hpp:

const int class_num=21; // for pascal voc caffemodel

[xiaozialei]

改成两类后C++的结果和demo.py运行的结果不一致，请问有朋友遇到过这个问题吗？