can matlab yolov2 detecotr integrated with android ?

hello

i just convert the yolo detection using mkl dnn example into c++ using matlab coder witch is only detect an object.

but faceing a lot of problem when i try to integrated with android studio using ndk and android.mk file

any help quick please

android.mk :

LOCAL_PATH := $(call my-dir)
#CLEAR ALL LOCAL VAR EXCIPT LOCAL VAR
include $(CLEAR_VARS)


OPENCV_LIB_TYPE:=SHARED
OPENCV_CAMERA_MODULES:=on
OPENCV_INSTALL_MODULES:=on

include D:/android/OpenCV-android-sdk/sdk/native/jni/OpenCV.mk

#native module
LOCAL_MODULE := vidTest
#c files
LOCAL_SRC_FILES := yolov2_detection/main.cpp \
					yolov2_detection/activations.cpp \
					yolov2_detection/cnn_api.cpp \
                    yolov2_detection/DeepLearningNetwork.cpp \
                    yolov2_detection/selectStrongestBboxMulticlass.cpp \
                    yolov2_detection/sortIdx.cpp \
                    yolov2_detection/sort.cpp \
                    yolov2_detection/yolov2_detection.cpp \
                    yolov2_detection/yolov2_detection_data.cpp \
                  	yolov2_detection/yolov2_detection_initialize.cpp \
                  	yolov2_detection/YOLOv2Network.cpp \
                  	yolov2_detection/MWYoloSoftmaxLayerImpl.cpp\
                    yolov2_detection/MWSigmoidLayerImpl.cpp\
                    yolov2_detection/MWMkldnnUtils.cpp\
                    yolov2_detection/MWSigmoidLayer.cpp\
                    yolov2_detection/rt_nonfinite.cpp\
                    yolov2_detection/MWFusedConvReLULayer.cpp\
                    yolov2_detection/MWFusedConvReLULayerImpl.cpp\
                    yolov2_detection/rtGetInf.cpp\
                    yolov2_detection/MWYoloExtractionLayerImpl.cpp\
                    yolov2_detection/MWYoloExtractionLayer.cpp\
                    yolov2_detection/rtGetNaN.cpp\
                    yolov2_detection/MWTargetNetworkImpl.cpp\
                    yolov2_detection/MWYoloSoftmaxLayer.cpp\
                    yolov2_detection/nullAssignment.cpp\
                    yolov2_detection/MWExponentialLayerImpl.cpp\
                    yolov2_detection/MWExponentialLayer.cpp\
                    yolov2_detection/MWElementwiseAffineLayerImpl.cpp\
                    yolov2_detection/MWElementwiseAffineLayer.cpp\
                    yolov2_detection/MWDepthConcatenationLayerImpl.cpp\
                    yolov2_detection/MWDepthConcatenationLayer.cpp\
                    yolov2_detection/MWConvLayerImpl.cpp\
                    yolov2_detection/MWConvLayer.cpp\
                    yolov2_detection/MWCNNLayerImpl.cpp\
                  	yolov2_detection/cnn_api.hpp





LOCAL_C_INCLUDES += D:/android/OpenCV-android-sdk/sdk/native/jni/include
LOCAL_LDLIBS     += -llog -ldl
LOCAL_CFLAGS    += -DOPENCV_OLDER_VISION
LOCAL_CPPFLAGS += -frtti
include $(BUILD_SHARED_LIBRARY)

this is the error

[Praveen Kumar Gajula]

Hi Sara,

Deep learning(DL) codegen using mkldnn is not plain C/C++ codegen which can be used on any platform.It has been tested and validated on dekstop,laptop for windows/Linux. It needs prebuilt mkldnn libraries which could be built as per the link.

If you want to deploy it on Android platforms you need to do DL codegen using arm-compute. Again this is not a plain C/C++ codegen it uses arm compute library for acceleration under the hood. So you need codegen + arm compute library to deploy on Android platform.

Hope this helps.

Thank you,

Praveen.

[sara alamri]

thanks

any link or documntation so i could assimble all things togather.

by codegen i used matlab coder is't ok ?

[Praveen Kumar Gajula]

We don't have link to explain about the workflow. It is more of an offline assmebling which you might have to do. There is no automated way to do that in MATLAB as of today . Please use the below links for android deployment manually.

To build arm compute library for android - Refer to this link

To do the codegen for arm compute use example - link

We are working on bringing the automated android deployment for deep learning networks.

Yes, MATLAB coder is the right toolbox for this work flow.

Thank you,

Praveen.

[sara alamri]

Thanks a lot One final question does it required ARM processor or it is ok with Intel processor in my laptop ?