topic OpticalFlowPyrLK performance in IntelĀ® Integrated Performance Primitives
https://community.intel.com/t5/Intel-Integrated-Performance/OpticalFlowPyrLK-performance/m-p/871227#M8947
We have executed the example 14-5 of the Manual Image Processing (vol.2) to test OpticalFlowPyrLK with the Yosemite (images of 252x316 pixels). We have used the following values for the input parameters:<BR />(1) A Gaussian kernel of length 5: 151, 21186, 5319, 2186, 151<BR />(2) Levels from 1 to 3 (numLevel)<BR />(3) rate =2<BR />(4) winSize of 3, 5 and 7<BR />(5) numIter of 1,2 and 3<BR />(6) Theshold=0<BR />(7) prevPt is initialized by using:<BR /> for (i=0;i<ROISIZE.WIDTH></ROISIZE.WIDTH> prevPt<I>.x=i%roiSize.width; prevPt<I>.y= i/roiSize.width;<BR /> }<BR />Then, the flow is obtained for all the pixels.<BR />(8) nextPt is equal than prevPt.<BR />We obtain 2879.2 cycles per pixel (in the input image) with a Gaussian filter of length 5, numLevel=1, numIter=1, winSize=3 (cycles of the processor clock obtained with instruction rdtsc). This is a high value. With other combination the number of cycles is yet higher. <BR />Nevertheless, with the IPPtest programps_ippcvem64t.exe we obtain 10 , 6, or 2 clocks per element (depends on the input parameters) for images of 256x256 pixels.<BR />What are we doing wrong? Is the source code for ps_ippcvem64t.exe available? We cannot reproduce the clocks per element of this test (we suppose clocks per element are cycles per pixel in this context).<BR /><BR />Thanks<BR /></I></I>Wed, 07 Jan 2009 18:06:54 GMTmancita2009-01-07T18:06:54ZOpticalFlowPyrLK performance
https://community.intel.com/t5/Intel-Integrated-Performance/OpticalFlowPyrLK-performance/m-p/871227#M8947
We have executed the example 14-5 of the Manual Image Processing (vol.2) to test OpticalFlowPyrLK with the Yosemite (images of 252x316 pixels). We have used the following values for the input parameters:<BR />(1) A Gaussian kernel of length 5: 151, 21186, 5319, 2186, 151<BR />(2) Levels from 1 to 3 (numLevel)<BR />(3) rate =2<BR />(4) winSize of 3, 5 and 7<BR />(5) numIter of 1,2 and 3<BR />(6) Theshold=0<BR />(7) prevPt is initialized by using:<BR /> for (i=0;i<ROISIZE.WIDTH></ROISIZE.WIDTH> prevPt<I>.x=i%roiSize.width; prevPt<I>.y= i/roiSize.width;<BR /> }<BR />Then, the flow is obtained for all the pixels.<BR />(8) nextPt is equal than prevPt.<BR />We obtain 2879.2 cycles per pixel (in the input image) with a Gaussian filter of length 5, numLevel=1, numIter=1, winSize=3 (cycles of the processor clock obtained with instruction rdtsc). This is a high value. With other combination the number of cycles is yet higher. <BR />Nevertheless, with the IPPtest programps_ippcvem64t.exe we obtain 10 , 6, or 2 clocks per element (depends on the input parameters) for images of 256x256 pixels.<BR />What are we doing wrong? Is the source code for ps_ippcvem64t.exe available? We cannot reproduce the clocks per element of this test (we suppose clocks per element are cycles per pixel in this context).<BR /><BR />Thanks<BR /></I></I>Wed, 07 Jan 2009 18:06:54 GMThttps://community.intel.com/t5/Intel-Integrated-Performance/OpticalFlowPyrLK-performance/m-p/871227#M8947mancita2009-01-07T18:06:54ZRe: OpticalFlowPyrLK performance
https://community.intel.com/t5/Intel-Integrated-Performance/OpticalFlowPyrLK-performance/m-p/871228#M8948
<DIV style="margin:0px;">Hi,<BR />IPPtest program ps_ippcvem64t.exe returns clocks per element (pixel) per feature point. So toreproduce it tou should devide your time by numFeat parameter.<BR /><BR />Igor</DIV>
<BR />Tue, 13 Jan 2009 08:11:48 GMThttps://community.intel.com/t5/Intel-Integrated-Performance/OpticalFlowPyrLK-performance/m-p/871228#M8948Igor_B_Intel12009-01-13T08:11:48Z