Home

SRCNN_OpenCV_GCC

• SRCNN is a forked version from shuwang's SRCNN_Cpp.

Performance

• SRCNN made to use OpenCV library by ShuWang, and it was enhanced performance with OpenMP.
• This performance measured in nVidia Jetson TX2 ( AARCH64 ), and only for 4 cores in active ( either disabled for as FPGA processing )
• All sources built with built-in OpenCV of nVidia sources, version 2.4.9.1
  • Non OpenMP: fresh measured as 41.1 ms
  • With OpenMP: fresh measured as 11.1 ms ( about 4 times faster than non-openmp )
  • libsrcnn ( non-OpenCV C++ native model ) : fresh measured as 4.5 ms ( about 3 times faster than opencv + openmp )

Notice,

• This project is not for non-OpenCV model, if you want to looking for more faster, more smaller, go for libsrcnn.