cyberfire / tensorflow-mtcnn Goto Github PK

Thanks for the code. how the model is freeze to mtcnn_frozen_model.pb?
I trained pnet, rnet and onets with tensoflow and the layer names things is quite different from yours and wondering how you freeze the model to a frozen .pb and .npy file.

I am using P5000 GPU with GPU memory of ~16GB.. This application uses 15.18GB as soon as it is launched!! is it expected?

I am working with standalone version of MTCNN C++ code...
Cuda : 9.0
CUDNN 7.5
Ubuntu 16.04

i am doing face recognition,using your c++ implement,but i want to test the time of under cpu and gpu,tensorflow defaultly use gpu ,and all the memory,you use the c api,i am not know which function to set the two value,i hope you can help me.Thank you!

hello,first thank you contribute your codes!
now i have a problem,originally,there are three models. How do you convert to the mtcnn_frozen_model.pb?
Is there have codes to convert?thanks!

Can I please know how to align face after detection in c++ as done in https://github.com/davidsandberg/facenet/blob/master/src/align/align_dataset_mtcnn.py

I have imported the c++ code into eclipse,
I am trying to run test.cpp , but I am unable to read .jpg images , though am able to read .png images

~/tensorflow-mtcnn-master/cpp/standalone$ sudo make test
the error is :
g++ test.o -o test tensorflow_mtcnn.o comm_lib.o utils.o -L/usr/local/lib -lopencv_shape -lopencv_stitching -lopencv_objdetect -lopencv_superres -lopencv_videostab -lopencv_calib3d -lopencv_features2d -lopencv_highgui -lopencv_videoio -lopencv_imgcodecs -lopencv_video -lopencv_photo -lopencv_ml -lopencv_imgproc -lopencv_flann -lopencv_core -Wl,-rpath,/usr/local/include/lib -L/usr/local/include/lib -ltensorflow
/usr/bin/ld: warning: libmklml_intel.so, needed by /usr/local/lib/libtensorflow.so, not found (try using -rpath or -rpath-link)
/usr/bin/ld: warning: libiomp5.so, needed by /usr/local/lib/libtensorflow.so, not found (try using -rpath or -rpath-link)
/usr/local/lib/libtensorflow.so：对‘dnnLRNCreateForward_F32’未定义的引用
/usr/local/lib/libtensorflow.so：对‘dnnConcatCreate_F32’未定义的引用
/usr/local/lib/libtensorflow.so：对‘dnnLRNCreateBackward_F32’未定义的引用
/usr/local/lib/libtensorflow.so：对‘dnnLayoutDelete_F32’未定义的引用
/usr/local/lib/libtensorflow.so：对‘dnnBatchNormalizationCreateBackward_v2_F32’未定义的引用
/usr/local/lib/libtensorflow.so：对‘cblas_dgemm’未定义的引用
/usr/local/lib/libtensorflow.so：对‘dnnPoolingCreateForward_F32’未定义的引用
/usr/local/lib/libtensorflow.so：对‘dnnLayoutDeserialize_F32’未定义的引用
/usr/local/lib/libtensorflow.so：对‘cblas_cgemm’未定义的引用
/usr/local/lib/libtensorflow.so：对‘dnnLayoutGetMemorySize_F32’未定义的引用
/usr/local/lib/libtensorflow.so：对‘dnnBatchNormalizationCreateForward_v2_F32’未定义的引用
/usr/local/lib/libtensorflow.so：对‘dnnExecute_F32’未定义的引用
/usr/local/lib/libtensorflow.so：对‘omp_in_parallel@VERSION’未定义的引用
/usr/local/lib/libtensorflow.so：对‘cblas_zgemm’未定义的引用
/usr/local/lib/libtensorflow.so：对‘dnnLayoutSerializationBufferSize_F32’未定义的引用
/usr/local/lib/libtensorflow.so：对‘dnnPoolingCreateBackward_F32’未定义的引用
/usr/local/lib/libtensorflow.so：对‘dnnConvolutionCreateBackwardBias_F32’未定义的引用
/usr/local/lib/libtensorflow.so：对‘dnnConvolutionCreateForward_F32’未定义的引用
/usr/local/lib/libtensorflow.so：对‘cblas_sgemm’未定义的引用
/usr/local/lib/libtensorflow.so：对‘dnnLayoutCompare_F32’未定义的引用
/usr/local/lib/libtensorflow.so：对‘dnnConvolutionCreateForwardBias_F32’未定义的引用
/usr/local/lib/libtensorflow.so：对‘dnnConversionExecute_F32’未定义的引用
//usr/local/lib/libtensorflow_framework.so：对‘i_calloc’未定义的引用
/usr/local/lib/libtensorflow.so：对‘omp_get_max_threads@VERSION’未定义的引用
/usr/local/lib/libtensorflow.so：对‘omp_get_thread_num@VERSION’未定义的引用
//usr/local/lib/libtensorflow_framework.so：对‘i_malloc’未定义的引用
/usr/local/lib/libtensorflow.so：对‘MKL_Domatcopy’未定义的引用
/usr/local/lib/libtensorflow.so：对‘GOMP_barrier@VERSION’未定义的引用
/usr/local/lib/libtensorflow.so：对‘dnnConvolutionCreateBackwardFilter_F32’未定义的引用
/usr/local/lib/libtensorflow.so：对‘omp_get_num_threads@VERSION’未定义的引用
//usr/local/lib/libtensorflow_framework.so：对‘i_free’未定义的引用
/usr/local/lib/libtensorflow.so：对‘MKL_Zomatcopy’未定义的引用
/usr/local/lib/libtensorflow.so：对‘dnnDelete_F32’未定义的引用
/usr/local/lib/libtensorflow.so：对‘dnnLayoutCreate_F32’未定义的引用
/usr/local/lib/libtensorflow.so：对‘dnnLayoutSerialize_F32’未定义的引用
/usr/local/lib/libtensorflow.so：对‘dnnLayoutCreateFromPrimitive_F32’未定义的引用
/usr/local/lib/libtensorflow.so：对‘MKL_Comatcopy’未定义的引用
/usr/local/lib/libtensorflow.so：对‘dnnSumCreate_F32’未定义的引用
/usr/local/lib/libtensorflow.so：对‘GOMP_parallel@VERSION’未定义的引用
/usr/local/lib/libtensorflow.so：对‘dnnConversionCreate_F32’未定义的引用
/usr/local/lib/libtensorflow.so：对‘MKL_Somatcopy’未定义的引用
/usr/local/lib/libtensorflow.so：对‘dnnReLUCreateBackward_F32’未定义的引用
/usr/local/lib/libtensorflow.so：对‘dnnReLUCreateForward_F32’未定义的引用
/usr/local/lib/libtensorflow.so：对‘dnnConvolutionCreateBackwardData_F32’未定义的引用
//usr/local/lib/libtensorflow_framework.so：对‘i_realloc’未定义的引用
collect2: error: ld returned 1 exit status
Makefile:45: recipe for target 'test' failed
make: *** [test] Error 1

how can i solve this problem?everybody give me recipe will thanks!!

I am using face detection using tensorflow 1.4 , but get below error while executing the code

/home/ashok/eclipseWorkspace/faceRecognition-x86_64_MTCNN/Libraries/tensorflow/include/tensorflow/core/lib/core/refcount.h:79] Check failed: ref_.load() == 0 (1 vs. 0)
generated from
Status run_status = sess->Run(input_tname,output_tname,output_node,&output_tensor);

form
run_PNet(std::unique_ptr<tensorflow::Session>& sess, cv::Mat& img, scale_window& win, std::vector<face_box>& box_list)
function
Please help resolve the issue

hello,i don't know the function of this code: working_img=working_img.t() ；why should do this? @cyberfire

Hi, cyberfire, Thanks for you codes. when I build the C++ version (tf_embedded) use bazel, there were some errors. It looks like that the tensorflow version dost not match well. I want to know which tensorflow version you used

./tensorflow/core/framework/tensor_shape.h:507: error: undefined reference to 'tensorflow::TensorShapeRep::DestructorOutOfLine()'
./tensorflow/core/public/session_options.h:28: error: undefined reference to 'tensorflow::ConfigProto::~ConfigProto()'
tensorflow/examples/mtcnn/main.cc:80: error: undefined reference to 'tensorflow::GraphDef::GraphDef()'
tensorflow/examples/mtcnn/main.cc:82: error: undefined reference to 'tensorflow::Env::Default()'
tensorflow/examples/mtcnn/main.cc:82: error: undefined reference to 'tensorflow::ReadBinaryProto(tensorflow::Env*, std::string const&, google::protobuf::MessageLite*)'
tensorflow/examples/mtcnn/main.cc:87: error: undefined reference to 'tensorflow::SessionOptions::SessionOptions()'
tensorflow/examples/mtcnn/main.cc:87: error: undefined reference to 'tensorflow::NewSession(tensorflow::SessionOptions const&)'
tensorflow/examples/mtcnn/main.cc:92: error: undefined reference to 'tensorflow::GraphDef::~GraphDef()'
tensorflow/examples/mtcnn/main.cc:92: error: undefined reference to 'tensorflow::GraphDef::~GraphDef()'
tensorflow/examples/mtcnn/main.cc:116: error: undefined reference to 'tensorflow::cpu_allocator()'
tensorflow/examples/mtcnn/main.cc:126: error: undefined reference to 'tensorflow::Tensor::Tensor(tensorflow::DataType, tensorflow::TensorShape const&, tensorflow::TensorBuffer*)'
./tensorflow/core/framework/tensor_shape.h:288: error: undefined reference to 'tensorflow::TensorShapeBasetensorflow::TensorShape::TensorShapeBase(tensorflow::gtl::ArraySlice)'
/usr/include/c++/4.9/bits/stl_pair.h:96: error: undefined reference to 'tensorflow::Tensor::~Tensor()'
tensorflow/examples/mtcnn/main.cc:205: error: undefined reference to 'cv::_OutputArray::_OutputArray(cv::Mat&)'
tensorflow/examples/mtcnn/main.cc:205: error: undefined reference to 'cv::_InputArray::_InputArray(cv::Mat const&)'
tensorflow/examples/mtcnn/main.cc:256: error: undefined reference to 'tensorflow::TensorShapeBasetensorflow::TensorShape::dim_size(int) const'
tensorflow/examples/mtcnn/main.cc:257: error: undefined reference to 'tensorflow::TensorShapeBasetensorflow::TensorShape::dim_size(int) const'
tensorflow/examples/mtcnn/main.cc:262: error: undefined reference to 'tensorflow::Tensor::tensor_data() const'
tensorflow/examples/mtcnn/main.cc:263: error: undefined reference to 'tensorflow::Tensor::tensor_data() const'
tensorflow/examples/mtcnn/main.cc:226: error: undefined reference to 'tensorflow::Tensor::~Tensor()'

Please review MTCNN TensorFlow code at https://github.com/look4pritam/TensorFlowMTCNN on CelebA branch.

since I want to use MTCNN for detection and alignment of another kind of object instead of face, I need to train on my own data. how to train on self-define data which has already been formatted same as FDDB?

Hello,

How can I add another .pb model that does face recognition and runs after ONet is run? All I was able to do is add it beneath the run_ONet function with its own session and graph:

run_ONet(sess,graph,working_img, rnet_boxes,total_onet_boxes);
	
//Our face recognition model
run_Ours(sess2, graph2,working_img, total_onet_boxes,total_our_boxes);
total_onet_boxes = total_our_boxes;

Hi,
I compiled the mtcnn embeded codes,then run it,got an error as below:
../../lib:/usr/local/lib:/usr/local/cuda-8.0/lib64:/usr/local/cuda-8.0/lib64:/usr/local/lib:
2017-12-27 11:02:44.944551: I tensorflow/core/platform/cpu_feature_guard.cc:137] Your CPU supports instructions that this TensorFlow binary was not compiled to use: SSE4.1 SSE4.2 AVX AVX2 FMA
2017-12-27 11:02:45.084213: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:895] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2017-12-27 11:02:45.084773: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1062] Found device 0 with properties:
name: GeForce GTX 970M major: 5 minor: 2 memoryClockRate(GHz): 1.038
pciBusID: 0000:01:00.0
totalMemory: 5.94GiB freeMemory: 5.36GiB
2017-12-27 11:02:45.084792: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1152] Creating TensorFlow device (/device:GPU:0) -> (device: 0, name: GeForce GTX 970M, pci bus id: 0000:01:00.0, compute capability: 5.2)
run PNet error
2017-12-27 11:02:45.539095: F ../../include/tensorflow/core/lib/core/refcount.h:79] Check failed: ref_.load() == 0 (1 vs. 0)
./run.sh: line 6: 9237 Aborted (core dumped) ./mtcnn_embeded

can you help me!
thanks !