Re: GStreamer 1.0 #241
Posted 7 November 2014 - 10:29
Patch for riff-media uploaded also on GStreamer bug for review.
@OpenPLi, do we want to use (master) git version for GStreamer bitbales or stable releases (eg 1.4.x)?
Unamed: 13E Quattro - 9E Quattro on IKUSI MS-0916
Re: GStreamer 1.0 #242
Re: GStreamer 1.0 #243
Posted 7 November 2014 - 10:35
Sebastian Dröge (slomo) [GStreamer developer] 2014-11-07 09:31:46 UTC (From update of attachment 290138 [details]) As said before this is not the right approach considering all the discussion we had so far. Your hardware probably needs different settings depending on the MPEG-4 Part 2 *profile* but not based on this magic fourccs. And mpeg4videoparse should already add the profile to the caps.So does profile works?
Unamed: 13E Quattro - 9E Quattro on IKUSI MS-0916
Re: GStreamer 1.0 #244
Re: GStreamer 1.0 #245
Posted 7 November 2014 - 11:23
Yeah, rejects tend to be quick
Let's try to get the issue completely clear and work to a solution from there.
- avi cannot store B frames (because they need to be decoded out-of-order which avi doesn't have provisioning for)
- divx (xvid???) creates an mpeg4-vc elementary stream that has B frames "packed" with other (P? I sounds unlikely) frames
- these "packed" frames are stored in avi like one "normal" frame
- the avi demuxer, on decoding, only produces I and P frames (where P frames may be "packed" frames including one or more (?) B frames)
- if you feed a "packed" frame to the decoder, it won't work (what happens actually, does it only decode the first frame or nothing at all?)
- so at some point between demuxing and decoding, the "packed" frame needs to "unpacked" into P and B frames, where also the PTS and DTS timestamps are regenerated
- does gstreamer have functionality for that anyway?
I guess the crux is to determine if this something that should be handled by the decoder or by the demuxer. I think it should be handled by the decoder (the way it is conceived, this way), but apparently our hardware decoder doesn't like it that way. Also I am not sure that at the point the elementary stream is fed to the decoder, there is still enough context for the decoder to determine if the "packed" frames are used, or that this information is stored in the (avi) container. If the only hint whether "packed" frames are used is in the (avi) fourcc, then I think we're on very thin ice. Isn't there any other way to determine this fact, from other metadata (either container or elementary stream)?
And in that case, I guess a conversion layer should be added between demuxer and decoder, which gstreamer can do, apparently?
Edited by Erik Slagter, 7 November 2014 - 11:23.
* Wavefrontier T90 with 28E/23E/19E/13E via SCR switches 2 x 2 x 6 user bands
I don't read PM -> if you have something to ask or to report, do it in the forum so others can benefit. I don't take freelance jobs.
Ik lees geen PM -> als je iets te vragen of te melden hebt, doe het op het forum, zodat anderen er ook wat aan hebben.
Re: GStreamer 1.0 #246
Posted 7 November 2014 - 13:47
<athoik> slomo: if mpeg4videoparse was working on avi files adding profile (and other informations) will it add also information about bitstream (packed / unpacked) ? <athoik> slomo: packed/unpacked avi bitstream http://itsjustonesandzeros.blogspot.com/2007/01/what-is-packed-bitstream.html <slomo> athoik: mpeg4videoparse should just unpack automatically <slomo> athoik: you will get exactly one frame per buffer <slomo> athoik: does that help? <athoik> slomo: we are doing the opposite! we are packing the unpack bitstream (and until now nobody knows why this is done) <slomo> athoik: well, that's a hardware specific thing then that you'll have to do in your sink :)
Edited by athoik, 7 November 2014 - 13:49.
Unamed: 13E Quattro - 9E Quattro on IKUSI MS-0916
Re: GStreamer 1.0 #247
Posted 7 November 2014 - 23:11
Until this get fixed we can use it to detect most xvid avi...
Subject: [PATCH] dvbvideosink: when mpeg4 doesnt have codec_data then handle it as xvid diff --git a/gstdvbvideosink.c b/gstdvbvideosink.c index d8d5635..f23cec7 100755 --- a/gstdvbvideosink.c +++ b/gstdvbvideosink.c @@ -1304,11 +1304,19 @@ static gboolean gst_dvbvideosink_set_caps(GstBaseSink *basesink, GstCaps *caps) { GST_INFO_OBJECT (self, "MPEG4 have codec data"); self->codec_data = gst_value_get_buffer(codec_data); - self->codec_type = CT_MPEG4_PART2; gst_buffer_ref (self->codec_data); + streamtype = 4; } - streamtype = 4; - GST_INFO_OBJECT (self, "MIMETYPE video/mpeg4 -> VIDEO_SET_STREAMTYPE, 4"); + else + { + /* HACK When mpeg4 doesn't have codec_data then most probably it's xvid */ + streamtype = 10; +#ifdef PACK_UNPACKED_XVID_DIVX5_BITSTREAM + self->must_pack_bitstream = TRUE; +#endif + } + self->codec_type = CT_MPEG4_PART2; + GST_INFO_OBJECT (self, "MIMETYPE video/mpeg4 -> VIDEO_SET_STREAMTYPE, %d", streamtype); } break; default: --When we have codec_data (after bug 739773 is fixed) then we need to find the magic bytes..
dvbvideosink gstdvbvideosink.c:1308:gst_dvbvideosink_set_caps:<dvbvideosink9> set_caps video/mpeg, mpegversion=(int)4, systemstream=(boolean)false, profile=(string)advanced-simple, level=(string)5, codec_data=(buffer)000001b0f5000001b5090000010000000120088685770307d31404221463000001b244697658353033623133393370000001b25876694430303634, width=(int)640, height=(int)272, framerate=(fraction)24000/1001, pixel-aspect-ratio=(fraction)1/1 codec_data=(buffer)000001b0f5000001b5090000010000000120088685770307d31404221463000001b2 44 69 76 58 353033623133393370000001b2 58 76 69 44 30303634xvid.png 19.09KB 5 downloads
Most probably we can already use codec_data to detect xvid files in mp4 format and use correct streamtype...
Attached Files
Unamed: 13E Quattro - 9E Quattro on IKUSI MS-0916
Re: GStreamer 1.0 #248
Posted 8 November 2014 - 14:17
@Erik
if you feed a "packed" frame to the decoder, it won't work (what happens actually, does it only decode the first frame or nothing at all?)
It doesn't work with streamtype 4, but works with streamtype 10
If the only hint whether "packed" frames are used is in the (avi) fourcc, then I think we're on very thin ice. Isn't there any other way to determine this fact, from other metadata (either container or elementary stream)?
AVI with fourcc XVID doesn't mean that packed frames are used, It only means that there is posibility of packed frames. So we set streamtype 10(handles packed frames) for decoder and pack unpacked bitstream on the fly, metadata about packed/unpacked are in bitstream.
if (sscanf((char*)data+pos, "DivX%d%c%d%cp", &tmp1, &c1, &tmp2, &c2) == 4 && (c1 == 'b' || c1 == 'B') && (c2 == 'p' || c2 == 'P')) { GST_INFO_OBJECT (self, "%s seen... already packed!", (char*)data+pos); self->must_pack_bitstream = FALSE; break; }
At least that's how I think it is.
@athoik
I don't think this is a bug. In dvbvideosink we don't require "profile" or "codec_data" field for mpeg4, so decodebin immediately chooses dvbvideosink instead of going through mpeg4videoparse. If we require "profile" field in dvbvideosink, it will go through mpeg4videoparse and it will append additional caps (codec_data, profile_field)
diff --git a/gstdvbvideosink.c b/gstdvbvideosink.c index d8d5635..fdd015d 100755 --- a/gstdvbvideosink.c +++ b/gstdvbvideosink.c @@ -178,10 +178,12 @@ GST_STATIC_PAD_TEMPLATE ( GST_STATIC_CAPS ( "video/mpeg, " #ifdef HAVE_MPEG4 - "mpegversion = (int) { 1, 2, 4 }, " -#else - "mpegversion = (int) { 1, 2 }, " + "mpegversion = (int) 4, " + "profile = (string) { simple, advanced-simple }, " + VIDEO_CAPS "; " #endif + "video/mpeg, " + "mpegversion = (int) { 1, 2 }, " VIDEO_CAPS "; " #ifdef HAVE_H264 "video/x-h264, "
gst-launch-1.0 -v playbin uri="file:///media/net/DOMA/Filmy/Unaveni slnkom/Unaveni-sluncem-2-Odpor-Cz.avi"|grep vbin /GstPlayBin:playbin0/GstPlaySink:playsink/GstBin:vbin/GstDVBVideoSink:dvbvideosink0: ts-offset = 0 /GstPlayBin:playbin0/GstPlaySink:playsink/GstBin:vbin.GstGhostPad:sink.GstProxyPad:proxypad10: caps = "video/mpeg\,\ mpegversion\=\(int\)4\,\ systemstream\=\(boolean\)false\,\ fourcc\=\(uint\)1145656920\,\ framerate\=\(fraction\)25/1\,\ width\=\(int\)720\,\ height\=\(int\)304\,\ parsed\=\(boolean\)true\,\ profile\=\(string\)advanced-simple\,\ level\=\(string\)5\,\ codec_data\=\(buffer\)000001b0f5000001b509000001000000012008868400670c5a1098518f\,\ pixel-aspect-ratio\=\(fraction\)1/1"
Re: GStreamer 1.0 #249
Posted 8 November 2014 - 17:39
Streamtypes??? Yet another avi aborration?It doesn't work with streamtype 4, but works with streamtype 10
AVI with fourcc XVID doesn't mean that packed frames are used, It only means that there is posibility of packed frames. So we set streamtype 10(handles packed frames) for decoder and pack unpacked bitstream on the fly, metadata about packed/unpacked are in bitstream.
So I am right in that the "fourcc" in itself doesn't signify "packed" format or not?
mpeg4-vc doesn't specify a "packed" format for it's elementary stream, so it will have to converted at some point before feeding it to the decoder, but I guess I've said that a few times before.
* Wavefrontier T90 with 28E/23E/19E/13E via SCR switches 2 x 2 x 6 user bands
I don't read PM -> if you have something to ask or to report, do it in the forum so others can benefit. I don't take freelance jobs.
Ik lees geen PM -> als je iets te vragen of te melden hebt, doe het op het forum, zodat anderen er ook wat aan hebben.
Re: GStreamer 1.0 #250
Re: GStreamer 1.0 #251
Posted 8 November 2014 - 20:41
When XviD packed then streamtype = 10
When XviD unpacked then streamtype = 10 and must_pack_bitstream = true
diff --git a/gstdvbvideosink.c b/gstdvbvideosink.c index d8d5635..a7ff0f7 100755 --- a/gstdvbvideosink.c +++ b/gstdvbvideosink.c @@ -150,6 +150,49 @@ void bitstream_put(struct bitstream *bit, unsigned long val, int bits) } #endif +gboolean is_xvid(size_t size, char *data, int *build) +{ + size_t pos = 0; + while (pos < size) + { + if (memcmp(&data[pos], "\x00\x00\x01\xb2", 4)) + { + pos++; + continue; + } + pos += 4; + if (size - pos < 8) break; + if (sscanf(data+pos, "XviD%d", build) == 1) + { + return TRUE; + } + } + return FALSE; +} + +gboolean is_xvid_packed(size_t size, char *data) +{ + int tmp1, tmp2; + unsigned char c1, c2; + size_t pos = 0; + while (pos < size) + { + if (memcmp(&data[pos], "\x00\x00\x01\xb2", 4)) + { + pos++; + continue; + } + pos += 4; + if (size - pos < 13) break; + if (sscanf(data+pos, "DivX%d%c%d%cp", &tmp1, &c1, &tmp2, &c2) == 4 && (c1 == 'b' || c1 == 'B') && (c2 == 'p' || c2 == 'P')) + { + return TRUE; + } + } + return FALSE; +} + + GST_DEBUG_CATEGORY_STATIC (dvbvideosink_debug); #define GST_CAT_DEFAULT dvbvideosink_debug @@ -176,12 +219,14 @@ GST_STATIC_PAD_TEMPLATE ( GST_PAD_SINK, GST_PAD_ALWAYS, GST_STATIC_CAPS ( - "video/mpeg, " #ifdef HAVE_MPEG4 - "mpegversion = (int) { 1, 2, 4 }, " -#else - "mpegversion = (int) { 1, 2 }, " + "video/mpeg, " + "mpegversion = (int) 4, " + "profile = (string) { simple, advanced-simple }, " + VIDEO_CAPS "; " #endif + "video/mpeg, " + "mpegversion = (int) { 1, 2 }, " VIDEO_CAPS "; " #ifdef HAVE_H264 "video/x-h264, " @@ -792,27 +837,10 @@ static GstFlowReturn gst_dvbvideosink_render(GstBaseSink *sink, GstBuffer *buffe } } - if (self->must_pack_bitstream) + if (self->must_pack_bitstream && is_xvid_packed(data_len, data)) { - int tmp1, tmp2; - unsigned char c1, c2; - unsigned int pos = 0; - while (pos < data_len) - { - if (memcmp(&data[pos], "\x00\x00\x01\xb2", 4)) - { - pos++; - continue; - } - pos += 4; - if (data_len - pos < 13) break; - if (sscanf((char*)data+pos, "DivX%d%c%d%cp", &tmp1, &c1, &tmp2, &c2) == 4 && (c1 == 'b' || c1 == 'B') && (c2 == 'p' || c2 == 'P')) - { - GST_INFO_OBJECT (self, "%s seen... already packed!", (char*)data+pos); - self->must_pack_bitstream = FALSE; - break; - } - } + GST_INFO_OBJECT (self, "Xvid: already packed, must_pack_bitstream = false"); + self->must_pack_bitstream = FALSE; } #endif @@ -1299,16 +1327,47 @@ static gboolean gst_dvbvideosink_set_caps(GstBaseSink *basesink, GstCaps *caps) break; case 4: { + streamtype = 4; + self->codec_type = CT_MPEG4_PART2; const GValue *codec_data = gst_structure_get_value(structure, "codec_data"); if (codec_data) { GST_INFO_OBJECT (self, "MPEG4 have codec data"); self->codec_data = gst_value_get_buffer(codec_data); - self->codec_type = CT_MPEG4_PART2; gst_buffer_ref (self->codec_data); +#if GST_VERSION_MAJOR >= 1 + GstMapInfo map; + gst_buffer_map(self->codec_data, &map, GST_MAP_READ); + gboolean parsed; + if(is_xvid(map.size, (char*)map.data, &mpegversion)) + { + streamtype = 10; + GST_INFO_OBJECT (self, "XviD: build %04d detected, using streamtype %d", mpegversion, streamtype); +#ifdef PACK_UNPACKED_XVID_DIVX5_BITSTREAM + if(is_xvid_packed(map.size, (char*)map.data)) + { + GST_INFO_OBJECT (self, "XviD: packed bitstream detected, must_pack_bitstream = false"); + self->must_pack_bitstream = FALSE; + } + else + { + GST_INFO_OBJECT (self, "XviD: unpacked bitstream detected, must_pack_bitstream = true"); + self->must_pack_bitstream = TRUE; + } +#endif + } + else if(gst_structure_get_boolean(structure, "parsed", &parsed) && parsed) + { + streamtype = 10; + GST_INFO_OBJECT (self, "XviD: parsed detected, using streamtype %d, must_pack_bitstream = true", streamtype); +#ifdef PACK_UNPACKED_XVID_DIVX5_BITSTREAM + self->must_pack_bitstream = TRUE; +#endif + } + gst_buffer_unmap(self->codec_data, &map); +#endif } - streamtype = 4; - GST_INFO_OBJECT (self, "MIMETYPE video/mpeg4 -> VIDEO_SET_STREAMTYPE, 4"); + GST_INFO_OBJECT (self, "MIMETYPE video/mpeg4 -> VIDEO_SET_STREAMTYPE, %d", streamtype); } break; default:Now that we get parsed (because we request profile in caps, and that is correct ofc) I cannot understand the behaviour.
When avi is parsed, I think it is removing packed bitstream...
Please have a look, I run out of ideas...
Attached Files
Unamed: 13E Quattro - 9E Quattro on IKUSI MS-0916
Re: GStreamer 1.0 #252
Posted 9 November 2014 - 14:32
Unpacked xvid avi play just fine with streamtype = 4 and ofc must_pack_bitstream = false!
So do we still need the hack?
PS. It takes only few seconds to unpack an xvid avi (transfer might take longer ), use unpackmp4 from here http://www.xmixdrix..../unpackmp4.html
Unamed: 13E Quattro - 9E Quattro on IKUSI MS-0916
Re: GStreamer 1.0 #253
Posted 9 November 2014 - 15:02
It would be nice if we can handle automatically both packed and unpacked xvid. How should we inform the user that they should unpack their videos? So we "only" need to know whether we have a packed/unpacked xvid. I already try to find a solution by looking into mediainfo source code.
Re: GStreamer 1.0 #254
Posted 9 November 2014 - 15:07
I suggest to look at the unpackmp4 source code (http://www.xmixdrix....packmp4-src.zip).
At least please verify that unpacked avi will work fine with gstreamer using streamtype = 4 (without any patches actually!, dts and pts patch is not required and with that applied it doesn't work)
Edited by athoik, 9 November 2014 - 15:08.
Unamed: 13E Quattro - 9E Quattro on IKUSI MS-0916
Re: GStreamer 1.0 #255
Posted 9 November 2014 - 15:18
Mediainfo scans the file and checks whether there are 2 frames in one block.
http://sourceforge.n...Mpeg4v.cpp#l559
So I think it's quite similar to unpackmp4 source.
Do we always get all data for one frame in gstdvbvideosink? Then we can scan the data whether there are 2 vops defined.
But I see a problem. First frame is a I frame. We have to decide immediately whether we use streamtype 4 or 10. I guess it makes problems when we write I frame with streamtype 4 and then switch to 10 because data is packed.
Or we need to buffer 2-3 frames in the sink
Edited by betacentauri, 9 November 2014 - 15:19.
Re: GStreamer 1.0 #256
Posted 9 November 2014 - 15:52
http://cgit.freedesk...deoparse.c#n627
Unamed: 13E Quattro - 9E Quattro on IKUSI MS-0916
Re: GStreamer 1.0 #257
Posted 9 November 2014 - 17:54
In caps we dont have the details required to detect packed avi, but in gst_dvbvideosink_render we do...
So if we haven't written the pts yet (first time that gst_dvbvideosink_render is called) and we have xvid with packed data, switch to streamtype = 10!
if (!self->pts_written) hexdump(data, data_len); int build; if (self->codec_type == CT_MPEG4_PART2 && !self->pts_written && is_xvid(data_len, data, &build) && is_xvid_packed(data_len, data)) { if(ioctl(self->fd, VIDEO_SET_STREAMTYPE, 10) < 0) GST_INFO_OBJECT (self, "XviD: packed detected, setting streamtype to 10 failed!!!"); else GST_INFO_OBJECT (self, "XviD: packed detected, setting streamtype to 10"); }but again packed xvid is not working
Attached Files
Unamed: 13E Quattro - 9E Quattro on IKUSI MS-0916
Re: GStreamer 1.0 #258
Re: GStreamer 1.0 #259
Re: GStreamer 1.0 #260
Posted 10 November 2014 - 20:34
Subject: [PATCH] dvbvideosink: add profile to caps and ignore dummy packed B-Frames diff --git a/gstdvbvideosink.c b/gstdvbvideosink.c index d8d5635..37b3ac0 100755 --- a/gstdvbvideosink.c +++ b/gstdvbvideosink.c @@ -176,12 +176,14 @@ GST_STATIC_PAD_TEMPLATE ( GST_PAD_SINK, GST_PAD_ALWAYS, GST_STATIC_CAPS ( - "video/mpeg, " #ifdef HAVE_MPEG4 - "mpegversion = (int) { 1, 2, 4 }, " -#else - "mpegversion = (int) { 1, 2 }, " + "video/mpeg, " + "mpegversion = (int) 4, " + "profile = (string) { simple, advanced-simple }, " + VIDEO_CAPS "; " #endif + "video/mpeg, " + "mpegversion = (int) { 1, 2 }, " VIDEO_CAPS "; " #ifdef HAVE_H264 "video/x-h264, " @@ -815,6 +817,9 @@ static GstFlowReturn gst_dvbvideosink_render(GstBaseSink *sink, GstBuffer *buffe } } #endif + /* remove dummy packed B-Frame */ + if (self->codec_type == CT_MPEG4_PART2 && data_len <= 7) + goto ok; pes_header[0] = 0; pes_header[1] = 0; --
PS. Try without DTS & PTS patch in PES header
Attached Files
Unamed: 13E Quattro - 9E Quattro on IKUSI MS-0916
Also tagged with one or more of these keywords: gstreamer, 1.0, openpli
DVB subtitles support in eServiceMP3/GStreamerStarted by DimitarCC, 17 Oct 2024 DVB, Subtitles, GStreamer |
|
|||
Change from openvix to openpli - lose existing hdd recordings?Started by xdoktor, 30 Dec 2023 openpli, openvix, hdd, recordings |
|
|||
Having Trouble While Installing This SoftwareStarted by CharleyDavis, 27 Jun 2023 OpenPLi |
|
|||
Faild to flash or update OPENPLIStarted by dede_one, 8 Oct 2022 openpli |
|
|||
hd+ funktioneret nichtStarted by JeppeG, 29 Sep 2022 Vu+, hd+, oscam, openpli |
|
4 user(s) are reading this topic
0 members, 3 guests, 0 anonymous users
-
Bing (1)