RE: [stereo-pipeline] Error during Low-Res Correlation

93 views
Skip to first unread message

Alexandrov, Oleg (ARC-TI)[SGT, INC]

unread,
Sep 29, 2017, 12:29:35 PM9/29/17
to Paul Montesano, stereo-...@lists.nasa.gov, ames-stereo-pi...@googlegroups.com
Paul,

The second error comes from the first. About the first, it depends on how you invoked ASP. Did you by chance use --stereo-algorithm 1 or 2? Did you use any non-default correlation kernel, or maybe one was read from your stereo.default?

I think if you either use --stereo-algorithm 0 or --corr-kernel 7 7 (or 3 or 5) it should be fine.

There is always a chance we mixed up something. If you give me the precise command, I will try to see if this error is intentional or not.

Oleg



From: stereo-pipeline [stereo-pipe...@lists.nasa.gov] on behalf of Paul Montesano [paul.mo...@gmail.com]
Sent: Friday, September 29, 2017 8:31 AM
To: stereo-...@lists.nasa.gov
Subject: [stereo-pipeline] Error during Low-Res Correlation

Hi Oleg,

I'm getting an error that I havent seen before that occurs during the generation of out-D_sub.tif

I''m pasting two error reports below.  

[1] The output to the prompt, which reports a 'vw::NoImpIErr' error
[2] The log file, which reports a fileio error

Could you provide me with a guess as to whether this suggests a filesystem issue of a software issue?

Thank you,
Paul

[1] Prompt
-------------
[ 2017-Sep-29 11:19:08 ] : Stage 1 --> LOW-RESOLUTION CORRELATION
Looking for IP file: /att/gpfsfs/briskfs01/ppl/pmontesa/outASP/WV03_20160703_104001001F044500_104001001EB92100/out-WV03_20160703_16JUL03155213-104001001F04__WV03_20160703_16JUL03155112-104001001EB9.match
        --> Using interest points to determine search window.
            * Locating Interest Points
Using 2329 interest points per tile (1024^2 px).
            Processing left image
            Processing right image
            Removing IP near nodata
            Building descriptors
            Found interest points:
              left: 8687
             right: 8967
        --> Matching interest points
           Matching: [***************************************************.] 99%
            Matched points: 711
        --> Homography: Matrix3x3((0.942282,0.00768316,25.9106)(-0.023823,1.00605,12.7002)(-4.34721e-05,4.08677e-05,1))
            * Writing match file: /att/gpfsfs/briskfs01/ppl/pmontesa/outASP/WV03_20160703_104001001F044500_104001001EB92100/out-L_sub__R_sub.match
            * Using cached match file: /att/gpfsfs/briskfs01/ppl/pmontesa/outASP/WV03_20160703_104001001F044500_104001001EB92100/out-L_sub__R_sub.match
        --> Detected search range: (Origin: (-860, -708) width: 1881 height: 1635)
D_sub search range: (Origin: (-95, -79) width: 204 height: 177) px
        --> Low-resolution disparity:[***.................................] 8%terminate called after throwing an instance of 'vw::NoImplErr'
  what():  Census transform is only available in size 3, 5, and 7!

Stereo step  failed

[2] Log file
----------
[ 2017-Sep-29 11:16:27 ] : Stage 1 --> LOW-RESOLUTION CORRELATION 
2017-09-29 11:16:27 {0} [ console ] : --> Using interest points to determine search window.
2017-09-29 11:16:27 {0} [ console ] :     * Locating Interest Points
2017-09-29 11:16:27 {0} [ console ] : Using 2331 interest points per tile (1024^2 px).
2017-09-29 11:16:27 {0} [ console ] :     Processing left image
2017-09-29 11:16:27 {0} [ console ] :     Processing right image
2017-09-29 11:16:28 {0} [ console ] :     Removing IP near nodata
2017-09-29 11:16:28 {0} [ console ] :     Building descriptors
2017-09-29 11:16:28 {0} [ console ] :     Found interest points:
      left: 7904
2017-09-29 11:16:28 {0} [ console ] :      right: 9151
2017-09-29 11:16:28 {0} [ console ] : --> Matching interest points
2017-09-29 11:16:29 {0} [ console ] : 
    Matched points: 1816
2017-09-29 11:16:30 {0} [ console ] : --> Homography: Matrix3x3((1.13981,0.0629956,-64.8719)(-0.0355666,1.27438,-125.175)(-2.76399e-05,0.00014104,1))
2017-09-29 11:16:30 {0} [ console ] :     * Writing match file: /att/gpfsfs/briskfs01/ppl/pmontesa/outASP/WV03_20150421_104001000A830500_104001000A487F00/out-L_sub__R_sub.match
2017-09-29 11:16:30 {0} [ console ] :     * Using cached match file: /att/gpfsfs/briskfs01/ppl/pmontesa/outASP/WV03_20150421_104001000A830500_104001000A487F00/out-L_sub__R_sub.match
2017-09-29 11:16:30 {0} [ console ] : --> Detected search range: (Origin: (-255, -284) width: 469 height: 664)
2017-09-29 11:16:30 {0} [ fileio ] : Error: GdalIO: `/att/gpfsfs/briskfs01/ppl/pmontesa/outASP/WV03_20150421_104001000A830500_104001000A487F00/out-D_sub.tif' not recognised as a supported file format.  (code = 4)
2017-09-29 11:16:31 {0} [ fileio ] : Error: GdalIO: Unable to open /att/gpfsfs/briskfs01/ppl/pmontesa/outASP/WV03_20150421_104001000A830500_104001000A487F00/out-D_sub.tif to obtain file list. (code = 4)

Alexandrov, Oleg (ARC-TI)[SGT, INC]

unread,
Sep 29, 2017, 12:41:00 PM9/29/17
to Paul Montesano, stereo-...@lists.nasa.gov, ames-stereo-pi...@googlegroups.com
Paul,

I would suggest you try --stereo-algorithm 2 as well. In my experience that one (MGM) can give less noisy results than the first (SGM), but it also depends on your input. 

I don't know how old your ASP is, those algorithms have had their memory usage improved a lot, so if you run into issues you can fetch the latest build from https://byss.arc.nasa.gov/stereopipeline/daily_build/

Oleg



From: Paul Montesano [paul.mo...@gmail.com]
Sent: Friday, September 29, 2017 9:36 AM
To: Alexandrov, Oleg (ARC-TI)[SGT, INC]; stereo-...@lists.nasa.gov
Cc: ames-stereo-pi...@googlegroups.com
Subject: Re: [stereo-pipeline] Error during Low-Res Correlation

Oleg,
That resolved the problem. Thank you.

I was using -stereo-algorithm 1 but hadn't switched to --corr-kernel 7 7

-Paul
Reply all
Reply to author
Forward
0 new messages