hansren1024 / face-tracking-using-cnn-and-optical-flow Goto Github PK

Great work!

but how to track multi face?

你好，请问你有尝试过用caffe实现该代码吗？ @HansRen1024

我发现该追踪由于是在cpu上实现的，对CPU的占用还是挺高的，请问是否有gpu版本的？

/usr/bin/ld: ../demo: hidden symbol `__cpu_model' in /usr/lib/gcc/x86_64-linux-gnu/5/libgcc.a(cpuinfo.o) is referenced by DSO
/usr/bin/ld: final link failed: Bad value
collect2: error: ld returned 1 exit status
CMakeFiles/demo.dir/build.make:114: recipe for target '../demo' failed
make[2]: *** [../demo] Error 1
CMakeFiles/Makefile2:105: recipe for target 'CMakeFiles/demo.dir/all' failed
make[1]: *** [CMakeFiles/demo.dir/all] Error 2
Makefile:83: recipe for target 'all' failed
make: *** [all] Error 2

Thanks for sharing your work with us.I have a few questions .
If removing the classifiers in the algorithm,the learning module in the original algorithm is gone, then use the tld is not much sense？I think the biggest contribution of TLD is the framework of tracking-learning-detection.
Whether the effect will be better,if i use another tracker,faster than tld？

Hi, to calculate the ratio between new length and old length, I think it needs a sqrt before the norm? As the norm calculates the sum of square value only. This issue exists also in the python version, thank you!

Can I replace mtcnn with another detector?

我uncomment了视频存储相关的语句，但是运行后目录内没有保存的文件，是哪里还需要调整嘛？

hi, I issued some questions again in https://blog.csdn.net/renhanchi/article/details/81668987, pls have a look and give some clues, thanks.

您好，请问下这个文件在哪里？谢谢

Thanks for your job! But when I test your code with a local movie, after function "tld.processFrame(last_gray,current_gray,pbox,status);" in ncnn_mtcnn_tld_so.cpp ,sometimes the x, width,height of pbox will be negative, and "mtcnn.rnet(img, pbox)" will be nan or -nan.

您好，请问下您make时gcc的版本号是多少？
我的是4.8.2，make时报错：
“/usr/include/c++/4.8.2/bits/stl_algo.h:2266:34: 错误：将类型为‘Bbox&’的引用初始化为类型为‘const Bbox’的表达式无效”
这应该是gcc的bug吧，查到说4.9.0+就不会出现这个问题了。
谢谢～

调试的时候发现，在tld跟踪的时候，尺度几乎是都是1，没有变化， 请问这是tld哪里出问题了吗？

MacOSX 不支持吗？ncnn这个库明确说明了，xcode do not support openmp feature，我就是在mac上用xcode编译的ncnn，然后在Face-Tracking-Based-on-OpenTLD-and-RNet/build编译的时候，就遇到了这个问题。

clang: error: unsupported option '-fopenmp'

Build for MacOSX
install xcode and protobuf

Because the compiler bundled with xcode do not support openmp feature, you cannot enable the multithreading inference feature of ncnn library, if you build with xcode.

$ homebrew install protobuf
$ cd
$ mkdir -p build
$ cd build
$ cmake ..
$ make -j4
$ make install
pick build/install folder for further usage

Ali:build admin$ make
Scanning dependencies of target TLD
[ 16%] Building CXX object CMakeFiles/TLD.dir/include/TLD.o
clang: error: unsupported option '-fopenmp'
make[2]: *** [CMakeFiles/TLD.dir/include/TLD.o] Error 1
make[1]: *** [CMakeFiles/TLD.dir/all] Error 2
make: *** [all] Error 2

can give some url, i want know what is TLD. thanks.

如题，大牛有什么好思路修改成多人脸跟踪么？谢谢