dafucoding / mtcnn_caffe Goto Github PK

int count = confidence->count()/2;
const float* confidence_data = confidence->cpu_data();
confidence_data += count;
const float* reg_data = reg->cpu_data();

Can I know why here confidence_data += count, but reg_data doesn't += count?

您好，我用这个代码测试了下test1,test2.jpg 速度并不快，约400-500ms 每张 。就算是将minsize改到30，40 也没有多少变化。 请问你写的20ms是多大尺寸的图片？

Hi, when I ran the demo, ./build/examples/MTSrc/MTMain.bin model_dir test_image.jpg
I met the problem of : "Segmentation fault"

RTX2080Ti 处理每帧图像 500ms 左右。而且运行一分钟左右就会报以下错误，请问可能是哪里的问题？

F1020 14:24:46.977366 18296 syncedmem.cpp:71] Check failed: error == cudaSuccess (2 vs. 0) out of memory

Or, how did you compile it?
Thanks.

change link below, if someone else has the same problam

MTCNN::MTCNN(const std::string &proto_model_dir){
//#ifdef CPU_ONLY
Caffe::set_mode(Caffe::CPU);
//#else
// Caffe::set_mode(Caffe::GPU);
//#endif

Hello,
Code worked fine for me, but after I changed something (contains this algorithm snippet), all of projects which contain this algorithm can't run successfully anymore. I'm pretty sure they worked fine before I debug something...

When I tried to construct MTCNN object, PNet_->num_inputs() returns millions inputs :<
I'm inexperienced in Python/Caffe, could you please help me out? I want to know what did I ruin.

Thanks!

我在Ubuntu下执行以下命令（路径已修改）时报“Segmentation fault (core dumped)”的错误
./build/examples/MTSrc/MTMain.bin '/home/dafu/workspace/MTCNN_Caffe/examples/MTmodel' '/home/dafu/workspace/MTCNN_Caffe/examples/MTSrc/test2.jpg'
gdb调试结果：
Thread 1 "MTMain.bin" received signal SIGSEGV, Segmentation fault.
0x00007ffff741c29f in caffe::ConvolutionLayer::Forward_gpu (this=0xb61f5b0, bottom=std::vector of length 1, capacity 1 = {...}, top=std::vector of length 1, capacity 1 = {...}) at src/caffe/layers/conv_layer.cu:22
22 std::cout<<*(tempTop+i)<<std::endl;
请问问题在哪呢？

您好，我想请问：Detect里为什么要对图像转置（ sample_single = sample_single.t();）？

Hi DaFu,

Thanks a lot for sharing your code. Did you get a chance to benchmark your code on CPU?

Thanks,
Sri

请问我想改成按照一个batch来测试，要怎么改呢？麻烦么？

Hi,
How did you train this MTmodel?
Did you change any settings compared to the training process in the raw paper?
Can you provide training code if u don't mind?

Hi!
Whether I can compile MTMain.cpp in other version CAFFE, and what new code you have added to the your caffe project?

Thank you!

Thanks for the great work.

Can this repo be used for training my own models?

请问为何在我的GTX1080Ti上测试450450像素的图像都需要500ms?，而且我已将试过了将下面 四行注释掉：
//const Dtype tempTop = top[0]->gpu_data();
//for(int i=0;i<10;i++){
//std::cout<<*(tempTop+i)<<std::endl;
//}

Detect 1080X1920F1229 22:31:31.232031 10620 data_transformer.cpp:167] Check fail
ed: channels == datum_channels (1095 vs. 3)
*** Check failure stack trace: ***

The MTMain.bin is the result of the MTMain.cpp file compiled output in Caffe.

Why does it take me 1 second to load caffe model on my 1050 card ，But more than a minute to load it on my V100 card

if #define CPU_ONLY the project work ok
if use gpu the work error at function MTCNN::Detect(const cv::Mat& image,std::vector& faceInfo,int minSize,double* threshold,double factor)
line PNet_->Forward();
thx

Forward_gpu() has:
// const Dtype* tempTop = top[0]->gpu_data();
// for(int i=0;i<10;i++){
// std::cout<<*(tempTop+i)<<std::endl;
// }

为啥会出现这几行代码？直接打印gpu上的数据没有报错？

这三个变量是留作备用的，表示头部姿态角度，现在还没有用到。

(py3.6_caffe) zhangxin@zhangxin-AW:~/github/MTCNN_Caffe$ sudo make
Package opencv3 was not found in the pkg-config search path.
Perhaps you should add the directory containing `opencv3.pc'
to the PKG_CONFIG_PATH environment variable
No package 'opencv3' found
PROTOC src/caffe/proto/caffe.proto
CXX .build_release/src/caffe/proto/caffe.pb.cc
CXX src/caffe/data_reader.cpp
In file included from ./include/caffe/util/device_alternate.hpp:40:0,
                 from ./include/caffe/common.hpp:19,
                 from src/caffe/data_reader.cpp:6:
./include/caffe/util/cudnn.hpp: In function ‘const char* cudnnGetErrorString(cudnnStatus_t)’:
./include/caffe/util/cudnn.hpp:21:10: warning: enumeration value ‘CUDNN_STATUS_RUNTIME_PREREQUISITE_MISSING’ not handled in switch [-Wswitch]
   switch (status) {
          ^
./include/caffe/util/cudnn.hpp:21:10: warning: enumeration value ‘CUDNN_STATUS_RUNTIME_IN_PROGRESS’ not handled in switch [-Wswitch]
./include/caffe/util/cudnn.hpp:21:10: warning: enumeration value ‘CUDNN_STATUS_RUNTIME_FP_OVERFLOW’ not handled in switch [-Wswitch]
./include/caffe/util/cudnn.hpp: In function ‘void caffe::cudnn::setConvolutionDesc(cudnnConvolutionStruct**, cudnnTensorDescriptor_t, cudnnFilterDescriptor_t, int, int, int, int)’:
./include/caffe/util/cudnn.hpp:113:70: error: too few arguments to function ‘cudnnStatus_t cudnnSetConvolution2dDescriptor(cudnnConvolutionDescriptor_t, int, int, int, int, int, int, cudnnConvolutionMode_t, cudnnDataType_t)’
       pad_h, pad_w, stride_h, stride_w, 1, 1, CUDNN_CROSS_CORRELATION));
                                                                      ^
./include/caffe/util/cudnn.hpp:15:28: note: in definition of macro ‘CUDNN_CHECK’
     cudnnStatus_t status = condition; \
                            ^~~~~~~~~
In file included from ./include/caffe/util/cudnn.hpp:5:0,
                 from ./include/caffe/util/device_alternate.hpp:40,
                 from ./include/caffe/common.hpp:19,
                 from src/caffe/data_reader.cpp:6:
/usr/local/cuda/include/cudnn.h:537:27: note: declared here
 cudnnStatus_t CUDNNWINAPI cudnnSetConvolution2dDescriptor( cudnnConvolutionDescriptor_t convDesc,
                           ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Makefile:575: recipe for target '.build_release/src/caffe/data_reader.o' failed
make: *** [.build_release/src/caffe/data_reader.o] Error 1

I solve it by

copy caffe/include/caffe/util/cudnn.hpp to ./include/caffe/util/cudnn.hpp

Hi,
I want to know the training data used in this project?