Harris Corner Detection¶

harris example resides in L2/examples/harris directory.

This benchmark tests the performance of harris function. The harris function detects corners in the image using harris corner detection and Non-maximum suppression algorithms.

The tutorial provides a step-by-step guide that covers commands for building and running kernel.

Executable Usage¶

Work Directory(Step 1)

The steps for library download and environment setup can be found in README file of L2 folder. For getting the design,

cd L2/example/harris

Build kernel(Step 2)

Run the following make command to build your XCLBIN and host binary targeting a specific device. Please be noticed that this process will take a long time, maybe couple of hours.

export OPENCV_INCLUDE=< path-to-opencv-include-folder >
export OPENCV_LIB=< path-to-opencv-lib-folder >
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:< path-to-opencv-lib-folder >
export DEVICE=< path-to-platform-directory >/< platform >.xpfm
make host xclbin TARGET=hw

Run kernel(Step 3)

To get the benchmark results, please run the following command.

make run TARGET=hw

Example output(Step 4)

-----------Harris Design---------------
INFO: Running OpenCL section.
Found Platform
Platform Name: Xilinx
XCLBIN File Name: krnl_harris
INFO: Importing Vitis_Libraries/vision/L2/examples/harris/Xilinx_Harris_L2_Test_vitis_hw_u200/build_dir.hw.xilinx_u200_xdma_201830_2/krnl_harris.xclbin
Loading: 'Vitis_Libraries/vision/L2/examples/harris/Xilinx_Harris_L2_Test_vitis_hw_u200/build_dir.hw.xilinx_u200_xdma_201830_2/krnl_harris.xclbin'
Kernel Created
Kernel Args set
Kernel called

Data copied from device to host
Execution done!
ocv corner count = 428, Hls corner count = 446
Commmon = 405         Success = 90.807175     Loss = 5.373832         Gain = 9.192825
Test Passed

------------------------------------------------------------

Profiling¶

The harris design is validated on Alveo u200 board at 300 MHz frequency. The hardware resource utilizations are listed in the following table.

Table 1 Hardware resources for Harris corner detection¶
Dataset			LUT	BRAM	FF	DSP
Resolution	NPPC	other params	LUT	BRAM	FF	DSP
4K	8	L1 Norm, Filter - 3x3	40460	227	27236	208
FHD	8	L1 Norm, Filter - 3x3	22478	113	16021	138

The performance is shown below

Table 2 Performance numbers in terms of FPS (Frames Per Second) for Harris corner detection¶
Dataset	FPS(CPU)	FPS(FPGA)
4k (3840x2160)	22	289
Full HD(1920x1080)	62	1100