How to speed up deep learning(i.mx6 quad plus)???

[gustcool2]

Hi.

I'm using i.mx6 quad plus processor and I'm a beginner.

I trained face(data) with google mobilenet ssd v3 object detection.

And, I installed opencv 4.3 on i.mx6 quad plus board.

It works well(face object detection), but too slow.

Speed is 6 frame/sec, even though only detect face.

I want to improve speed.(at leaset 15frame/sec)

Is it possible to speed up using the I.MX6's GPU?

And how?

[igorpadykov]

Hi JunTae

one can try to follow below link, also additional document was sent via mail

Computer Vision on i.MX Processors: Building OpenCV-2.4.X for Freescale's i.MX6 BSP (Yocto) 

Best regards
igor
-----------------------------------------------------------------------------------------------------------------------
Note: If this post answers your question, please click the Correct Answer button. Thank you!
-----------------------------------------------------------------------------------------------------------------------

[gustcool2]

Hi igor

Thanks for your answer.

But I have to use opencv>=4.3 for mobilenet ssd v3.

Is there any other way to install opecv 4.3 on yocto?

I just compiled opencv-4.3 and install using usb

Here are opencv cmake configuration.

======================================================================================

set(GCC_COMPILER_VERSION "" CACHE STRING "GCC Compiler version")
set(GNU_MACHINE "arm-poky-linux-gnueabi" CACHE STRING "GNU compiler triple")

set(CMAKE_SYSTEM_NAME linux)
set(CMAKE_SYSTEM_PROCESSOR arm)
set(MY_SYSROOT /opt/fsl-imx-fb/3.14.52-1.1.1/sysroots/cortexa9hf-vfp-neon-poky-linux-gnueabi CACHE PATH "ARM cross compilation system root")
set(MY_CROSS_DIR /opt/fsl-imx-fb/3.14.52-1.1.1/sysroots/i686-pokysdk-linux/usr/bin/arm-poky-linux-gnueabi CACHE PATH "ARM cross compiler path")
set(CMAKE_C_COMPILER ${MY_CROSS_DIR}/arm-poky-linux-gnueabi-gcc)
set(CMAKE_CXX_COMPILER ${MY_CROSS_DIR}/arm-poky-linux-gnueabi-g++)
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} --sysroot=${MY_SYSROOT} -march=armv7-a -mfpu=neon -mfloat-abi=hard -mcpu=cortex-a9" CACHE INTERNAL ""FORCE)
set(CMAKE_CXX_FLAGS "${CMAKE_C_FLAGS} --sysroot=${MY_SYSROOT} -march=armv7-a -mfpu=neon -mfloat-abi=hard -mcpu=cortex-a9" CACHE INTERNAL ""FORCE)

include("${CMAKE_CURRENT_LIST_DIR}/arm.toolchain.cmake")

======================================================================================