Out of Order Timestamps

[Wes Miller]

I have a buffer probe on the src pad of a capsfilter.  In the callback I print out each buffer's timestamp.  The data is demuxed from a matroska contaner file and represents a recorded mjpeg

Here is a sample of the listing where x is a counter incremented as each new frame is processed:


     x = 1482  --->  4237392448l
     x = 1483  --->  4269392448l
     x = 1484  --->  7425152l
     x = 1485  --->  40425152l
     x = 1486  --->  75425152l
     x = 1487  --->  208425152l
     x = 1488  --->  240425152l
     x = 1489  --->  274425152l
     x = 1490  --->  309425152l

Are the frames really this badly out of order?  Are the timestamps reliable?  Is there a pipe element that I need to add after the demuxer to reorder the frames?

Wes

[Sudarshan Bisht]

Hi ,

 

It would be better if you print timestamp is GST_TIME_FORMAT. 

On Tue, Jun 8, 2010 at 10:28 PM, Wes Miller <[hidden email]> wrote:

I have a buffer probe on the src pad of a capsfilter.  In the callback I
print out each buffer's timestamp.  The data is demuxed from a matroska
contaner file and represents a recorded mjpeg

Here is a sample of the listing where x is a counter incremented as each new
frame is processed:


    x = 1482  --->  4237392448l
    x = 1483  --->  4269392448l
    x = 1484  --->  7425152l
    x = 1485  --->  40425152l
    x = 1486  --->  75425152l
    x = 1487  --->  208425152l
    x = 1488  --->  240425152l
    x = 1489  --->  274425152l
    x = 1490  --->  309425152l

Are the frames really this badly out of order?  Are the timestamps reliable?
Is there a pipe element that I need to add after the demuxer to reorder the
frames?

Wes

--
View this message in context: http://gstreamer-devel.966125.n4.nabble.com/Ouyt-of-Order-Timestamps-tp2247716p2247716.html
Sent from the GStreamer-devel mailing list archive at Nabble.com.

------------------------------------------------------------------------------
ThinkGeek and WIRED's GeekDad team up for the Ultimate
GeekDad Father's Day Giveaway. ONE MASSIVE PRIZE to the
lucky parental unit.  See the prize list and enter to win:
http://p.sf.net/sfu/thinkgeek-promo
_______________________________________________
gstreamer-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/gstreamer-devel

--
Regards,

Sudarshan Bisht

------------------------------------------------------------------------------
ThinkGeek and WIRED's GeekDad team up for the Ultimate
GeekDad Father's Day Giveaway. ONE MASSIVE PRIZE to the
lucky parental unit.  See the prize list and enter to win:
http://p.sf.net/sfu/thinkgeek-promo
_______________________________________________
gstreamer-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/gstreamer-devel

[Tim-Philipp Müller-2]

On Tue, 2010-06-08 at 09:58 -0700, Wes Miller wrote:

> I have a buffer probe on the src pad of a capsfilter.  In the callback I
> print out each buffer's timestamp.  The data is demuxed from a matroska
> contaner file and represents a recorded mjpeg
>
> Here is a sample of the listing where x is a counter incremented as each new
> frame is processed:

>      x = 1482  --->  4237392448l
>      x = 1483  --->  4269392448l
>      x = 1484  --->  7425152l
>      x = 1485  --->  40425152l
>      x = 1486  --->  75425152l
>      x = 1487  --->  208425152l
> (snip)
> Are the frames really this badly out of order?  Are the timestamps reliable?
> Is there a pipe element that I need to add after the demuxer to reorder the
> frames?

MJPEG streams should not have out-of-order timestamps. Two things to
check:

1. Are the timestamps in the file really what you think?
   Maybe you are just printing the timestamps incorrectly?

   gst-launch-0.10 -v filesink location=... ! matroskademux name=d \
          d.video_0 ! fakesink

2. Are feeding buffers with good timestamps into
   matroskamux in the first place? Check by looking
   at the debug log, or add an identity element into
   a gst-launch pipeline and use gst-launch -v to
   make it print the timestamps of buffers passing
   through.

Cheers
 -Tim



------------------------------------------------------------------------------
ThinkGeek and WIRED's GeekDad team up for the Ultimate
GeekDad Father's Day Giveaway. ONE MASSIVE PRIZE to the
lucky parental unit.  See the prize list and enter to win:
http://p.sf.net/sfu/thinkgeek-promo
_______________________________________________
gstreamer-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/gstreamer-devel

[Wes Miller]

>> Hi ,
 
>> It would be better if you print timestamp is GST_TIME_FORMAT.



Ok, here is the same set of samples in the requested format.  

C-code reads:

     g_print( "x = %d  --->  "GST_TIME_FORMAT" \n",  x,  
                  GST_TIME_ARGS( GST_BUFFER_TIMESTAMP( buffer  )));

     x = 1482  --->  u:00:00.000000055
     x = 1483  --->  u:00:00.000000055
     x = 1484  --->  u:00:00.000000055
     x = 1485  --->  u:00:00.000000055
     x = 1486  --->  u:00:00.000000055
     x = 1487  --->  u:00:00.000000056
     x = 1488  --->  u:00:00.000000056
     x = 1489  --->  u:00:00.000000056
     x = 1490  --->  u:00:00.000000056
     x = 1491  --->  u:00:00.000000056
     x = 1492  --->  u:00:00.000000056


Ok, they're in order, apparently, but this output is still weird;

The saved video is 1 minute in length and the printed timestamps show there are about 27 frames per second as expected.  However, why is the rightmost field of the timestamps showing seconds?  Shouldn't those be nanoseconds?

Wes

Wes

[Edward Hervey]

On Tue, 2010-06-08 at 11:35 -0700, Wes Miller wrote:
>
>      g_print( "x = %d  --->  "GST_TIME_FORMAT" \n",  x,  
>                   GST_TIME_ARGS( GST_BUFFER_TIMESTAMP( buffer  )));

Correct version:"x = %d  --->  %"GST_TIME_FORMAT" \n"


------------------------------------------------------------------------------
ThinkGeek and WIRED's GeekDad team up for the Ultimate
GeekDad Father's Day Giveaway. ONE MASSIVE PRIZE to the
lucky parental unit.  See the prize list and enter to win:
http://p.sf.net/sfu/thinkgeek-promo
_______________________________________________
gstreamer-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/gstreamer-devel

[Wes Miller]

Edward,

Thanks for the proofreading/lesson.  I didn't quite understand the doc to mean I'd need the "%".  My timestamps are correct now.

Tim,

thanks for the reminder on using identity.  Thanks to Edward Hervey and Sudarshan Bisht for the additional help.  Everything is correct now.

Correct format was:

      g_print( "x = %d  ---> % "GST_TIME_FORMAT" \n",
                            x,             GST_TIME_ARGS( GST_BUFFER_TIMESTAMP( buffer )));

Wes