diff mbox

[FFmpeg-devel] ffmpeg: Add spherical_mapping command-line option.

Message ID CAA0c1bCRKoZbO-GCEwGKvwEc-b7d3K9A1Oo6pBJ1CDAwMvsmBg@mail.gmail.com
State New
Headers show

Commit Message

Aaron Colwell Oct. 6, 2017, 3:19 p.m. UTC
Allows spherical mapping metadata to be injected into files.

Comments

Mark Thompson Oct. 6, 2017, 3:52 p.m. UTC | #1
On 06/10/17 16:19, Aaron Colwell wrote:
> Allows spherical mapping metadata to be injected into files.
> 
> From 6a86e9766708b9b74e4ae0ec6928a81df4041afc Mon Sep 17 00:00:00 2001
> From: Aaron Colwell <acolwell@google.com>
> Date: Fri, 6 Oct 2017 08:14:15 -0700
> Subject: [PATCH] ffmpeg: Add spherical_mapping command-line option.
> 
> Allows spherical mapping metadata to be injected into files.
> ---
>  fftools/ffmpeg.c     |  12 +++-
>  fftools/ffmpeg.h     |   6 ++
>  fftools/ffmpeg_opt.c | 160 ++++++++++++++++++++++++++++++++++++++++++++++++++-
>  3 files changed, 176 insertions(+), 2 deletions(-)

Perhaps I'm missing something about this patch, but I'm not seeing why this should be added as a set of special options to ffmpeg.c.  A bsf in libavcodec which edits or replaces the packet side data would seem to be a more generally useful way to achieve the same result.  (E.g. it would also be usable in other applications, and it would be able to edit it on input as well as output.)

- Mark
Aaron Colwell Oct. 6, 2017, 4:10 p.m. UTC | #2
Hi Mark,

You aren't missing anything. I haven't really used bsf's so it didn't occur
to me to use them. I'll take a look. Thanks for the suggestion.

Aaron

On Fri, Oct 6, 2017 at 8:53 AM Mark Thompson <sw@jkqxz.net> wrote:

> On 06/10/17 16:19, Aaron Colwell wrote:
> > Allows spherical mapping metadata to be injected into files.
> >
> > From 6a86e9766708b9b74e4ae0ec6928a81df4041afc Mon Sep 17 00:00:00 2001
> > From: Aaron Colwell <acolwell@google.com>
> > Date: Fri, 6 Oct 2017 08:14:15 -0700
> > Subject: [PATCH] ffmpeg: Add spherical_mapping command-line option.
> >
> > Allows spherical mapping metadata to be injected into files.
> > ---
> >  fftools/ffmpeg.c     |  12 +++-
> >  fftools/ffmpeg.h     |   6 ++
> >  fftools/ffmpeg_opt.c | 160
> ++++++++++++++++++++++++++++++++++++++++++++++++++-
> >  3 files changed, 176 insertions(+), 2 deletions(-)
>
> Perhaps I'm missing something about this patch, but I'm not seeing why
> this should be added as a set of special options to ffmpeg.c.  A bsf in
> libavcodec which edits or replaces the packet side data would seem to be a
> more generally useful way to achieve the same result.  (E.g. it would also
> be usable in other applications, and it would be able to edit it on input
> as well as output.)
>
> - Mark
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel@ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>
Mark Thompson Oct. 6, 2017, 4:35 p.m. UTC | #3
On 06/10/17 17:10, Aaron Colwell wrote:
> On Fri, Oct 6, 2017 at 8:53 AM Mark Thompson <sw@jkqxz.net> wrote:
>> On 06/10/17 16:19, Aaron Colwell wrote:
>>> Allows spherical mapping metadata to be injected into files.
>>>
>>> From 6a86e9766708b9b74e4ae0ec6928a81df4041afc Mon Sep 17 00:00:00 2001
>>> From: Aaron Colwell <acolwell@google.com>
>>> Date: Fri, 6 Oct 2017 08:14:15 -0700
>>> Subject: [PATCH] ffmpeg: Add spherical_mapping command-line option.
>>>
>>> Allows spherical mapping metadata to be injected into files.
>>> ---
>>>  fftools/ffmpeg.c     |  12 +++-
>>>  fftools/ffmpeg.h     |   6 ++
>>>  fftools/ffmpeg_opt.c | 160
>> ++++++++++++++++++++++++++++++++++++++++++++++++++-
>>>  3 files changed, 176 insertions(+), 2 deletions(-)
>>
>> Perhaps I'm missing something about this patch, but I'm not seeing why
>> this should be added as a set of special options to ffmpeg.c.  A bsf in
>> libavcodec which edits or replaces the packet side data would seem to be a
>> more generally useful way to achieve the same result.  (E.g. it would also
>> be usable in other applications, and it would be able to edit it on input
>> as well as output.)
> 
> You aren't missing anything. I haven't really used bsf's so it didn't occur
> to me to use them. I'll take a look. Thanks for the suggestion.

Having looked a little further I suspect this doesn't quite work right now.  The output BSFs do not run on the first packet before the header is written, and AVCodecParameters does not carry the global side data - that means the edited side data won't be available at the right moment for global headers (it would be available per-packet, but the actual formats need it per-stream).  I'm not sure how hard that would be to fix.

- Mark
diff mbox

Patch

From 6a86e9766708b9b74e4ae0ec6928a81df4041afc Mon Sep 17 00:00:00 2001
From: Aaron Colwell <acolwell@google.com>
Date: Fri, 6 Oct 2017 08:14:15 -0700
Subject: [PATCH] ffmpeg: Add spherical_mapping command-line option.

Allows spherical mapping metadata to be injected into files.
---
 fftools/ffmpeg.c     |  12 +++-
 fftools/ffmpeg.h     |   6 ++
 fftools/ffmpeg_opt.c | 160 ++++++++++++++++++++++++++++++++++++++++++++++++++-
 3 files changed, 176 insertions(+), 2 deletions(-)

diff --git a/fftools/ffmpeg.c b/fftools/ffmpeg.c
index 1d248bc269..8c35090b9e 100644
--- a/fftools/ffmpeg.c
+++ b/fftools/ffmpeg.c
@@ -3092,6 +3092,10 @@  static int init_output_stream_streamcopy(OutputStream *ost)
             const AVPacketSideData *sd_src = &ist->st->side_data[i];
             uint8_t *dst_data;
 
+            if ((sd_src->type == AV_PKT_DATA_SPHERICAL && ost->spherical_mapping_overridden) ||
+                (sd_src->type == AV_PKT_DATA_STEREO3D && ost->stereo3d_overridden))
+                continue;
+
             dst_data = av_stream_new_side_data(ost->st, sd_src->type, sd_src->size);
             if (!dst_data)
                 return AVERROR(ENOMEM);
@@ -3528,7 +3532,13 @@  static int init_output_stream(OutputStream *ost, char *error, int error_len)
             int i;
             for (i = 0; i < ist->st->nb_side_data; i++) {
                 AVPacketSideData *sd = &ist->st->side_data[i];
-                uint8_t *dst = av_stream_new_side_data(ost->st, sd->type, sd->size);
+                uint8_t *dst;
+
+                if ((sd->type == AV_PKT_DATA_SPHERICAL && ost->spherical_mapping_overridden) ||
+                    (sd->type == AV_PKT_DATA_STEREO3D && ost->stereo3d_overridden))
+                    continue;
+
+                dst = av_stream_new_side_data(ost->st, sd->type, sd->size);
                 if (!dst)
                     return AVERROR(ENOMEM);
                 memcpy(dst, sd->data, sd->size);
diff --git a/fftools/ffmpeg.h b/fftools/ffmpeg.h
index f6c76bcc55..91ca6c2926 100644
--- a/fftools/ffmpeg.h
+++ b/fftools/ffmpeg.h
@@ -45,6 +45,8 @@ 
 #include "libavutil/hwcontext.h"
 #include "libavutil/pixfmt.h"
 #include "libavutil/rational.h"
+#include "libavutil/spherical.h"
+#include "libavutil/stereo3d.h"
 #include "libavutil/threadmessage.h"
 
 #include "libswresample/swresample.h"
@@ -237,6 +239,8 @@  typedef struct OptionsContext {
     int        nb_time_bases;
     SpecifierOpt *enc_time_bases;
     int        nb_enc_time_bases;
+    SpecifierOpt *spherical_mappings;
+    int        nb_spherical_mappings;
 } OptionsContext;
 
 typedef struct InputFilter {
@@ -487,6 +491,8 @@  typedef struct OutputStream {
     int top_field_first;
     int rotate_overridden;
     double rotate_override_value;
+    int spherical_mapping_overridden;
+    int stereo3d_overridden;
 
     AVRational frame_aspect_ratio;
 
diff --git a/fftools/ffmpeg_opt.c b/fftools/ffmpeg_opt.c
index 100fa76e46..4bc5104ce5 100644
--- a/fftools/ffmpeg_opt.c
+++ b/fftools/ffmpeg_opt.c
@@ -40,6 +40,8 @@ 
 #include "libavutil/parseutils.h"
 #include "libavutil/pixdesc.h"
 #include "libavutil/pixfmt.h"
+#include "libavutil/spherical.h"
+#include "libavutil/stereo3d.h"
 
 #define DEFAULT_PASS_LOGFILENAME_PREFIX "ffmpeg2pass"
 
@@ -1585,12 +1587,158 @@  static void check_streamcopy_filters(OptionsContext *o, AVFormatContext *oc,
     }
 }
 
+static int set_spherical_mapping(const char* opt, OutputStream *ost) {
+    typedef struct {
+        const AVClass* spherical_class;
+        int projection;
+
+        double yaw;
+        double pitch;
+        double roll;
+
+        int64_t bound_left;
+        int64_t bound_top;
+        int64_t bound_right;
+        int64_t bound_bottom;
+
+        int64_t padding;
+
+        int stereo_mode;
+    } SphericalMappingContext;
+
+#define OFFSET(x) offsetof(SphericalMappingContext, x)
+#define FLAGS AV_OPT_FLAG_VIDEO_PARAM
+
+    static const AVOption opts[] = {
+        { "projection", "projection", OFFSET(projection), AV_OPT_TYPE_INT,
+            { .i64 = -1 }, -1, AV_SPHERICAL_EQUIRECTANGULAR_TILE, FLAGS, "projection" },
+        { "equirectangular", "equirectangular projection", 0, AV_OPT_TYPE_CONST,
+            { .i64 = AV_SPHERICAL_EQUIRECTANGULAR }, INT_MIN, INT_MAX, FLAGS, "projection" },
+        { "cubemap", "cubemap projection", 0, AV_OPT_TYPE_CONST,
+            { .i64 = AV_SPHERICAL_CUBEMAP }, INT_MIN, INT_MAX, FLAGS, "projection" },
+        { "equirectangular_tile", "tiled equirectangular projection", 0, AV_OPT_TYPE_CONST,
+            { .i64 = AV_SPHERICAL_EQUIRECTANGULAR_TILE }, INT_MIN, INT_MAX, FLAGS, "projection" },
+        { "yaw", "initial yaw orientation in degrees", OFFSET(yaw), AV_OPT_TYPE_DOUBLE,
+            { .dbl = 0.0 }, -180.0, 180.0, FLAGS },
+        { "pitch", "initial pitch orientation in degrees", OFFSET(pitch), AV_OPT_TYPE_DOUBLE,
+            { .dbl = 0.0 }, -90.0, 90.0, FLAGS },
+        { "roll", "initial roll orientation in degrees", OFFSET(roll), AV_OPT_TYPE_DOUBLE,
+            { .dbl = 0.0 }, -180.0, 180.0, FLAGS },
+        { "bound_left", "tiled equirectangular left bound", OFFSET(bound_left), AV_OPT_TYPE_INT64,
+            { .i64 = 0 }, 0, UINT_MAX, FLAGS },
+        { "bound_top", "tiled equirectangular top bound", OFFSET(bound_top), AV_OPT_TYPE_INT64,
+            { .i64 = 0 }, 0, UINT_MAX, FLAGS },
+        { "bound_right", "tiled equirectangular right bound", OFFSET(bound_right), AV_OPT_TYPE_INT64,
+            { .i64 = 0 }, 0, UINT_MAX, FLAGS },
+        { "bound_bottom", "tiled equirectangular bottom bound", OFFSET(bound_bottom), AV_OPT_TYPE_INT64,
+            { .i64 = 0 }, 0, UINT_MAX, FLAGS },
+        { "padding", "cubemap padding in pixels", OFFSET(padding), AV_OPT_TYPE_INT64,
+            { .i64 = 0 }, 0, UINT_MAX, FLAGS },
+        { "stereo_mode", "stereo_mode", OFFSET(stereo_mode), AV_OPT_TYPE_INT,
+            { .i64 = -1 }, -1, AV_STEREO3D_TOPBOTTOM, FLAGS, "stereo_mode" },
+        { "top-bottom", "Top/Bottom stereo mode", 0, AV_OPT_TYPE_CONST,
+            { .i64 = AV_STEREO3D_TOPBOTTOM }, INT_MIN, INT_MAX, FLAGS, "stereo_mode" },
+        { "left-right", "Left/Right stereo mode", 0, AV_OPT_TYPE_CONST,
+            { .i64 = AV_STEREO3D_SIDEBYSIDE }, INT_MIN, INT_MAX, FLAGS, "stereo_mode" },
+        { NULL }
+    };
+#undef OFFSET
+#undef FLAGS
+
+    static const AVClass spherical_mapping_class = {
+        .class_name = "",
+        .item_name  = av_default_item_name,
+        .option     = opts,
+        .version    = LIBAVUTIL_VERSION_INT,
+    };
+
+    SphericalMappingContext ctx = {
+        .spherical_class = &spherical_mapping_class
+    };
+
+    AVStream* st = ost->st;
+    size_t spherical_mapping_size = 0;
+    AVSphericalMapping *spherical_mapping = NULL;
+    int ret;
+
+    if (!opt)
+        return AVERROR(EINVAL);
+
+    av_opt_set_defaults(&ctx);
+    ret = av_set_options_string(&ctx, opt, "=", ",");
+    if (ret < 0)
+        return ret;
+
+    if (ctx.projection == -1) {
+        av_log(NULL, AV_LOG_ERROR, "projection must be specified\n");
+        return AVERROR(EINVAL);
+    }
+
+    if (ctx.padding > 0 && ctx.projection != AV_SPHERICAL_CUBEMAP) {
+        av_log(NULL, AV_LOG_ERROR, "padding only allowed for AV_SPHERICAL_CUBEMAP projection.\n");
+        return AVERROR(EINVAL);
+    }
+
+    if ((ctx.bound_left > 0 || ctx.bound_top > 0 || ctx.bound_right > 0 ||
+         ctx.bound_bottom > 0) && ctx.projection != AV_SPHERICAL_EQUIRECTANGULAR_TILE) {
+        av_log(NULL, AV_LOG_ERROR, "bounds only allowed for AV_SPHERICAL_EQUIRECTANGULAR_TILE projection.\n");
+        return AVERROR(EINVAL);
+    }
+
+    spherical_mapping = av_spherical_alloc(&spherical_mapping_size);
+    if (!spherical_mapping)
+        return AVERROR(ENOMEM);
+
+    spherical_mapping->projection = (enum AVSphericalProjection)ctx.projection;
+    spherical_mapping->yaw = (int32_t)(ctx.yaw * (1 << 16));
+    spherical_mapping->pitch = (int32_t)(ctx.pitch * (1 << 16));
+    spherical_mapping->roll = (int32_t)(ctx.roll * (1 << 16));
+
+    if (ctx.projection == AV_SPHERICAL_CUBEMAP) {
+        spherical_mapping->padding = (uint32_t)ctx.padding;
+    } else if (ctx.projection == AV_SPHERICAL_EQUIRECTANGULAR_TILE) {
+        spherical_mapping->bound_left = (uint32_t)ctx.bound_left;
+        spherical_mapping->bound_top = (uint32_t)ctx.bound_top;
+        spherical_mapping->bound_right = (uint32_t)ctx.bound_right;
+        spherical_mapping->bound_bottom = (uint32_t)ctx.bound_bottom;
+    }
+
+    if (ret >= 0) {
+        ret = av_stream_add_side_data(st, AV_PKT_DATA_SPHERICAL, spherical_mapping, spherical_mapping_size);
+
+        if (ret < 0) {
+            av_freep(&spherical_mapping);
+        }
+
+        ost->spherical_mapping_overridden = 1;
+
+        if (ctx.stereo_mode != -1) {
+          AVStereo3D* stereo3d = av_stereo3d_alloc();
+
+          if (!stereo3d)
+            return AVERROR(ENOMEM);
+
+          stereo3d->type = (enum AVStereo3DType)(ctx.stereo_mode);
+          ret = av_stream_add_side_data(st, AV_PKT_DATA_STEREO3D, stereo3d, sizeof(*stereo3d));
+
+          if (ret < 0) {
+            av_freep(&stereo3d);
+            return ret;
+          }
+
+          ost->stereo3d_overridden = 1;
+        }
+    }
+
+    return ret;
+}
+
 static OutputStream *new_video_stream(OptionsContext *o, AVFormatContext *oc, int source_index)
 {
     AVStream *st;
     OutputStream *ost;
     AVCodecContext *video_enc;
-    char *frame_rate = NULL, *frame_aspect_ratio = NULL;
+    char *frame_rate = NULL, *frame_aspect_ratio = NULL, *spherical_mapping = NULL;
 
     ost = new_output_stream(o, oc, AVMEDIA_TYPE_VIDEO, source_index);
     st  = ost->st;
@@ -1617,6 +1765,12 @@  static OutputStream *new_video_stream(OptionsContext *o, AVFormatContext *oc, in
 
     MATCH_PER_STREAM_OPT(filter_scripts, str, ost->filters_script, oc, st);
     MATCH_PER_STREAM_OPT(filters,        str, ost->filters,        oc, st);
+    MATCH_PER_STREAM_OPT(spherical_mappings, str, spherical_mapping, oc, st);
+
+    if (spherical_mapping && set_spherical_mapping(spherical_mapping, ost) < 0) {
+        av_log(NULL, AV_LOG_FATAL, "Invalid spherical_mapping: %s\n", spherical_mapping);
+        exit_program(1);
+    }
 
     if (!ost->stream_copy) {
         const char *p = NULL;
@@ -3642,6 +3796,10 @@  const OptionDef options[] = {
         "automatically insert correct rotate filters" },
     { "hwaccel_lax_profile_check", OPT_BOOL | OPT_EXPERT,                        { &hwaccel_lax_profile_check},
         "attempt to decode anyway if HW accelerated decoder's supported profiles do not exactly match the stream" },
+    { "spherical_mapping", OPT_VIDEO | HAS_ARG  | OPT_STRING | OPT_SPEC |
+                           OPT_OUTPUT,                                           { .off = OFFSET(spherical_mappings) },
+        "set spherical mapping for video stream", "spherical_mapping" },
+
 
     /* audio options */
     { "aframes",        OPT_AUDIO | HAS_ARG  | OPT_PERFILE | OPT_OUTPUT,           { .func_arg = opt_audio_frames },
-- 
2.14.2.920.gcf0c67979c-goog