Menu
▾
▴
Re: [Teem-users] Sum of all values in a volume
|
From: Chourasia, A. <am...@sd...> - 2013-10-03 15:51:51
|
Hi Gordon, A fresh checkout from new repo address worked https://svn.code.sf.net/p/teem/code/teem/trunk While I was updating my earlier checkout version with older address, some stuff may not be updating, hence the issue. None the less thanks for helping out. Cheers -A On Oct 3, 2013, at 2:26 AM, Gordon L. Kindlmann wrote: > Hi Amit, > > You've run into the problem that sourceforge continues to serve some version, an old and un-update-able version, from https://teem.svn.sourceforge.net/svnroot/teem/teem/trunk teem > but you really should only be using, as SF said, > https://svn.code.sf.net/p/teem/code/teem/trunk > > Try checking that code out and building it. With that, when I do your command, I get: > > min: 0 > max: 4.0885190722395799e-11 > > Or with this: > unu make -h -t double -s 393 892 23 -i dump.bin \ > | unu axmerge -a 0 1 \ > | unu project -a 0 -m sum \ > | unu save -f text > > 6.1351656e-08 > > Let me know if you get something different. > > Gordon > > On Oct 2, 2013, at 4:07 PM, Chourasia, Amit wrote: > >> David: Nice, that works. >> >> Gordon: unu minmax is not reporting correct result for a binary data with doubles >> Here is an example >> >> unu make -h -t double -s 393 892 23 -i dump.bin \ >> | unu minmax - >> min: 0 >> max: 0.000000 >> >> Whereas the correct result should be >>> Min = 0 >>> Max = 4.08852e-11 >> >> >> The data dump.bin is available here >> http://users.sdsc.edu/~amit/forums/dump.bin >> >> Could this be a bug or my teem compilation is borked (In cmake configuration I see that its linking to 64 bit libraries.) >> >> -A >> >> >> >> >> >> On Oct 2, 2013, at 1:45 PM, David Weinstein wrote: >> >>> Hi Amit, >>> >>> Another solution, which will work with older version of Teems, is to insert a last stub axis before projecting: >>> >>> unu make -t float -s 1000000 -i vol.bin \ >>> | unu axinsert -a 1 \ >>> | unu project -a 0 -m sum \ >>> | unu save -f text >>> >>> Cheers, >>> Dave >>> >>> >>> On Oct 2, 2013, at 10:56 AM, "Chourasia, Amit" <am...@sd...> wrote: >>> >>>> Gordon, >>>> >>>> I checked out revision 6124, via >>>> %svn co https://teem.svn.sourceforge.net/svnroot/teem/teem/trunk teem >>>> >>>> I still get the error with unu project that it cannot sum 1-D array >>>> % unu make -t float -s 1000000 -i vol.bin \ >>>> | unu project -a 0 -m sum -i - >>>> (_nrrdFormatNRRD_read: reading raw data ... done) >>>> unu project: error projecting nrrd: >>>> [nrrd] nrrdProject: sorry, currently need at least 2-D array to project >>>> >>>> Could you tell me which version has this capability and how to fetch it? >>>> >>>> House keeping note: I get the following warning with SVN perhaps make this change on the webpage >>>> svn: Repository moved permanently to 'https://svn.code.sf.net/p/teem/code/teem/trunk'; please relocate >>>> >>>> I have the following work around, but a full unu solution like you mentioned will be perfect >>>> unu make -t float -s 10000 100 -i vol.bin \ >>>> | unu project -a 0 -m sum -i - \ >>>> | unu save -f text \ >>>> | paste -sd + - \ >>>> | bc >>>> >>>> BTW, thanks for reminding the -h option, one less pipe can save a lot of time for large data. >>>> >>>> Cheers >>>> --Amit >>>> >>>> >>>> >>>> On Oct 2, 2013, at 10:19 AM, Gordon L. Kindlmann wrote: >>>> >>>>> Hi Amit, >>>>> >>>>> As of some recent revision on the SVN trunk, you can unu project 1D arrays, and get out a singleton 1-element 1-D array. >>>>> >>>>> Also, you can use unu axmerge to lose the raster structure, so for your data, you could: >>>>> >>>>> unu make -h -t float -s 64 64 64 -i blob \ >>>>> | unu axmerge -a 0 1 \ >>>>> | unu project -a 0 -m sum \ >>>>> | unu save -f text >>>>> >>>>> Note the -h on "unu make"; it delays the actual read of blob and saves you one piping of data. Probably a minor difference but might help with larger data. >>>>> >>>>> But if all you care about is the sum, and you're creating the nrrd as needed (with unu make), you could also just: >>>>> >>>>> unu make -h -t float -s 262144 -i blob | unu project -a 0 -m sum | unu save -f text >>>>> >>>>> I'm using "unu save -f text" here as a way of getting out a textual representation, but that's only a 32-bit float in ASCII; you could also save it (the 1-element 1-D array) to a .nrrd file to record the value exactly. Or you could pipe to "unu minmax" which does an exact ASCII representation of a 64-bit double (via printf with %.17g). >>>>> >>>>> Gordon >>>>> >>>>> On Oct 1, 2013, at 2:25 PM, Chourasia, Amit wrote: >>>>> >>>>>> Hi, >>>>>> >>>>>> Is there a way to find the sum of all values in a volume? >>>>>> I thought this method may be there in unu 1op, but its not. >>>>>> >>>>>> Say I have a binary floating point 3D scalar data as "blob" . >>>>>> I could use project and sum along axes to get a 1d data, but is there a way to sum 1D data using teem. >>>>>> >>>>>> unu make -t float -s 64 64 64 -i blob | \ >>>>>> unu project -a 0 -m sum -i - | \ >>>>>> unu project -a 0 -m sum -i - >>>>>> ??? >>>>>> >>>>>> I would appreciate any pointers. >>>>>> >>>>>> Thanks >>>>>> -Amit >>>>>> >>>>>> PS: I am writing a very short documentation for quickly inspecting binary data using teem. >>>>>> ------------------------------------------------------------------------------ >>>>>> October Webinars: Code for Performance >>>>>> Free Intel webinars can help you accelerate application performance. >>>>>> Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most from >>>>>> the latest Intel processors and coprocessors. See abstracts and register > >>>>>> http://pubads.g.doubleclick.net/gampad/clk?id=60134791&iu=/4140/ostg.clktrk >>>>>> _______________________________________________ >>>>>> teem-users mailing list >>>>>> tee...@li... >>>>>> https://lists.sourceforge.net/lists/listinfo/teem-users >>>>> >>>> >>>> >>>> ------------------------------------------------------------------------------ >>>> October Webinars: Code for Performance >>>> Free Intel webinars can help you accelerate application performance. >>>> Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most from >>>> the latest Intel processors and coprocessors. See abstracts and register > >>>> http://pubads.g.doubleclick.net/gampad/clk?id=60134791&iu=/4140/ostg.clktrk >>>> _______________________________________________ >>>> teem-users mailing list >>>> tee...@li... >>>> https://lists.sourceforge.net/lists/listinfo/teem-users >>> >> >> >> ------------------------------------------------------------------------------ >> October Webinars: Code for Performance >> Free Intel webinars can help you accelerate application performance. >> Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most from >> the latest Intel processors and coprocessors. See abstracts and register > >> http://pubads.g.doubleclick.net/gampad/clk?id=60134791&iu=/4140/ostg.clktrk >> _______________________________________________ >> teem-users mailing list >> tee...@li... >> https://lists.sourceforge.net/lists/listinfo/teem-users > |