Compute and set VP9 Level if it is not already set

The VP9 level is computed when the container is missing a codec config or if the level is missing from the codec config. This fixes VP9 in ISO-BMFF files generated by FFmpeg v4.0.2 or earlier which does not have level set in the codec config. Fixes #469. Change-Id: I685bfd48be16ee6b2209da1c3173f7d6bb02b36a
shaka-project · Sep 13, 2018 · d0978b3 · d0978b3
1 parent 0709db4
commit d0978b3
Show file tree

Hide file tree

Showing 18 changed files with 182 additions and 16 deletions.
diff --git a/packager/app/test/testdata/opus-vp9-mp4-with-encryption/bear-320x240-vp9-opus-video.mp4 b/packager/app/test/testdata/opus-vp9-mp4-with-encryption/bear-320x240-vp9-opus-video.mp4
diff --git a/...pp/test/testdata/opus-vp9-mp4-with-encryption/decrypted-bear-320x240-vp9-opus-video-0.mp4 b/...pp/test/testdata/opus-vp9-mp4-with-encryption/decrypted-bear-320x240-vp9-opus-video-0.mp4
diff --git a/packager/app/test/testdata/opus-vp9-mp4-with-encryption/output.mpd b/packager/app/test/testdata/opus-vp9-mp4-with-encryption/output.mpd
@@ -20,7 +20,7 @@
       <ContentProtection schemeIdUri="urn:uuid:1077efec-c0b2-4d02-ace3-3c1e52e2fb4b">
         <cenc:pssh>AAAANHBzc2gBAAAAEHfv7MCyTQKs4zweUuL7SwAAAAExMjM0NTY3ODkwMTIzNDU2AAAAAA==</cenc:pssh>
       </ContentProtection>
-      <Representation id="1" bandwidth="231745" codecs="vp09.00.10.08.01.02.02.02.00" mimeType="video/mp4" sar="427:320">
+      <Representation id="1" bandwidth="231745" codecs="vp09.00.20.08.01.02.02.02.00" mimeType="video/mp4" sar="427:320">
         <BaseURL>bear-320x240-vp9-opus-video.mp4</BaseURL>
         <SegmentBase indexRange="1063-1130" timescale="1000000" presentationTimeOffset="37000">
           <Initialization range="0-1062"/>

diff --git a/packager/app/test/testdata/vp9-webm-with-blockgroup/bear-vp9-blockgroup-video.webm b/packager/app/test/testdata/vp9-webm-with-blockgroup/bear-vp9-blockgroup-video.webm
diff --git a/packager/app/test/testdata/vp9-webm-with-blockgroup/output.mpd b/packager/app/test/testdata/vp9-webm-with-blockgroup/output.mpd
@@ -3,7 +3,7 @@
 <MPD xmlns="urn:mpeg:dash:schema:mpd:2011" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:xlink="http://www.w3.org/1999/xlink" xsi:schemaLocation="urn:mpeg:dash:schema:mpd:2011 DASH-MPD.xsd" xmlns:cenc="urn:mpeg:cenc:2013" profiles="urn:mpeg:dash:profile:isoff-on-demand:2011" minBufferTime="PT2S" type="static" mediaPresentationDuration="PT2.7360000610351562S">
   <Period id="0">
     <AdaptationSet id="0" contentType="video" width="320" height="240" frameRate="1000000/33000" subsegmentAlignment="true" par="4:3">
-      <Representation id="0" bandwidth="196039" codecs="vp09.00.10.08.01.02.02.02.00" mimeType="video/webm" sar="1:1">
+      <Representation id="0" bandwidth="196039" codecs="vp09.00.20.08.01.02.02.02.00" mimeType="video/webm" sar="1:1">
         <BaseURL>bear-vp9-blockgroup-video.webm</BaseURL>
         <SegmentBase indexRange="302-320" timescale="1000000">
           <Initialization range="0-301"/>

diff --git a/packager/app/test/testdata/vp9-webm/bear-320x240-vp9-opus-video.webm b/packager/app/test/testdata/vp9-webm/bear-320x240-vp9-opus-video.webm
diff --git a/packager/app/test/testdata/vp9-webm/output.mpd b/packager/app/test/testdata/vp9-webm/output.mpd
@@ -12,7 +12,7 @@
       </Representation>
     </AdaptationSet>
     <AdaptationSet id="1" contentType="video" width="320" height="240" frameRate="1000000/34000" subsegmentAlignment="true" par="16:9">
-      <Representation id="1" bandwidth="225727" codecs="vp09.00.10.08.01.02.02.02.00" mimeType="video/webm" sar="427:320">
+      <Representation id="1" bandwidth="225727" codecs="vp09.00.20.08.01.02.02.02.00" mimeType="video/webm" sar="427:320">
         <BaseURL>bear-320x240-vp9-opus-video.webm</BaseURL>
         <SegmentBase indexRange="302-350" timescale="1000000" presentationTimeOffset="37000">
           <Initialization range="0-301"/>

diff --git a/packager/app/test/testdata/webm-subsample-encryption/bear-320x180-vp9-altref-video.webm b/packager/app/test/testdata/webm-subsample-encryption/bear-320x180-vp9-altref-video.webm
diff --git a/...pp/test/testdata/webm-subsample-encryption/decrypted-bear-320x180-vp9-altref-video-0.webm b/...pp/test/testdata/webm-subsample-encryption/decrypted-bear-320x180-vp9-altref-video-0.webm
diff --git a/packager/app/test/testdata/webm-subsample-encryption/output.mpd b/packager/app/test/testdata/webm-subsample-encryption/output.mpd
@@ -6,7 +6,7 @@
       <ContentProtection schemeIdUri="urn:uuid:1077efec-c0b2-4d02-ace3-3c1e52e2fb4b" cenc:default_KID="31323334-3536-3738-3930-313233343536">
         <cenc:pssh>AAAANHBzc2gBAAAAEHfv7MCyTQKs4zweUuL7SwAAAAExMjM0NTY3ODkwMTIzNDU2AAAAAA==</cenc:pssh>
       </ContentProtection>
-      <Representation id="0" bandwidth="184009" codecs="vp09.00.10.08.00.02.02.02.00" mimeType="video/webm" sar="1:1">
+      <Representation id="0" bandwidth="184009" codecs="vp09.00.11.08.00.02.02.02.00" mimeType="video/webm" sar="1:1">
         <BaseURL>bear-320x180-vp9-altref-video.webm</BaseURL>
         <SegmentBase indexRange="353-371" timescale="1000000">
           <Initialization range="0-352"/>

diff --git a/...ager/app/test/testdata/webm-vp9-full-sample-encryption/bear-320x180-vp9-altref-video.webm b/...ager/app/test/testdata/webm-vp9-full-sample-encryption/bear-320x180-vp9-altref-video.webm
diff --git a/...t/testdata/webm-vp9-full-sample-encryption/decrypted-bear-320x180-vp9-altref-video-0.webm b/...t/testdata/webm-vp9-full-sample-encryption/decrypted-bear-320x180-vp9-altref-video-0.webm
diff --git a/packager/app/test/testdata/webm-vp9-full-sample-encryption/output.mpd b/packager/app/test/testdata/webm-vp9-full-sample-encryption/output.mpd
@@ -6,7 +6,7 @@
       <ContentProtection schemeIdUri="urn:uuid:1077efec-c0b2-4d02-ace3-3c1e52e2fb4b" cenc:default_KID="31323334-3536-3738-3930-313233343536">
         <cenc:pssh>AAAANHBzc2gBAAAAEHfv7MCyTQKs4zweUuL7SwAAAAExMjM0NTY3ODkwMTIzNDU2AAAAAA==</cenc:pssh>
       </ContentProtection>
-      <Representation id="0" bandwidth="184009" codecs="vp09.00.10.08.00.02.02.02.00" mimeType="video/webm" sar="1:1">
+      <Representation id="0" bandwidth="184009" codecs="vp09.00.11.08.00.02.02.02.00" mimeType="video/webm" sar="1:1">
         <BaseURL>bear-320x180-vp9-altref-video.webm</BaseURL>
         <SegmentBase indexRange="353-371" timescale="1000000">
           <Initialization range="0-352"/>

diff --git a/packager/media/codecs/vp_codec_configuration_record.cc b/packager/media/codecs/vp_codec_configuration_record.cc
@@ -52,6 +52,79 @@ void MergeField(const std::string& name,
   }
 }
 
+enum VP9Level {
+  LEVEL_UNKNOWN = 0,
+  LEVEL_1 = 10,
+  LEVEL_1_1 = 11,
+  LEVEL_2 = 20,
+  LEVEL_2_1 = 21,
+  LEVEL_3 = 30,
+  LEVEL_3_1 = 31,
+  LEVEL_4 = 40,
+  LEVEL_4_1 = 41,
+  LEVEL_5 = 50,
+  LEVEL_5_1 = 51,
+  LEVEL_5_2 = 52,
+  LEVEL_6 = 60,
+  LEVEL_6_1 = 61,
+  LEVEL_6_2 = 62,
+  LEVEL_MAX = 255
+};
+
+struct VP9LevelCharacteristics {
+  uint64_t max_luma_sample_rate;
+  uint32_t max_luma_picture_size;
+  double max_avg_bitrate;
+  double max_cpb_size;
+  double min_compression_ratio;
+  uint8_t max_num_column_tiles;
+  uint32_t min_altref_distance;
+  uint8_t max_ref_frame_buffers;
+};
+
+struct VP9LevelDefinition {
+  VP9Level level;
+  VP9LevelCharacteristics characteristics;
+};
+
+VP9Level LevelFromCharacteristics(uint64_t luma_sample_rate,
+                                  uint32_t luma_picture_size) {
+  // https://www.webmproject.org/vp9/levels/.
+  const VP9LevelDefinition vp9_level_definitions[] = {
+      {LEVEL_1, {829440, 36864, 200, 400, 2, 1, 4, 8}},
+      {LEVEL_1_1, {2764800, 73728, 800, 1000, 2, 1, 4, 8}},
+      {LEVEL_2, {4608000, 122880, 1800, 1500, 2, 1, 4, 8}},
+      {LEVEL_2_1, {9216000, 245760, 3600, 2800, 2, 2, 4, 8}},
+      {LEVEL_3, {20736000, 552960, 7200, 6000, 2, 4, 4, 8}},
+      {LEVEL_3_1, {36864000, 983040, 12000, 10000, 2, 4, 4, 8}},
+      {LEVEL_4, {83558400, 2228224, 18000, 16000, 4, 4, 4, 8}},
+      {LEVEL_4_1, {160432128, 2228224, 30000, 18000, 4, 4, 5, 6}},
+      {LEVEL_5, {311951360, 8912896, 60000, 36000, 6, 8, 6, 4}},
+      {LEVEL_5_1, {588251136, 8912896, 120000, 46000, 8, 8, 10, 4}},
+      {LEVEL_5_2, {1176502272, 8912896, 180000, 90000, 8, 8, 10, 4}},
+      {LEVEL_6, {1176502272, 35651584, 180000, 90000, 8, 16, 10, 4}},
+      {LEVEL_6_1, {2353004544u, 35651584, 240000, 180000, 8, 16, 10, 4}},
+      {LEVEL_6_2, {4706009088u, 35651584, 480000, 360000, 8, 16, 10, 4}},
+  };
+
+  for (const VP9LevelDefinition& def : vp9_level_definitions) {
+    // All the characteristic fields except max_luma_sample_rate and
+    // max_luma_picture_size are ignored to avoid the extra complexities of
+    // computing those values. It may result in incorrect level being returned.
+    // If this is a problem, please file a bug to
+    // https://github.com/google/shaka-packager/issues.
+    if (luma_sample_rate <= def.characteristics.max_luma_sample_rate &&
+        luma_picture_size <= def.characteristics.max_luma_picture_size) {
+      return def.level;
+    }
+  }
+
+  LOG(WARNING) << "Cannot determine VP9 level for luma_sample_rate ("
+               << luma_sample_rate << ") or luma_picture_size ("
+               << luma_picture_size << "). Returning LEVEL_1.";
+  return LEVEL_1;
+}
+
 }  // namespace
 
 VPCodecConfigurationRecord::VPCodecConfigurationRecord() {}
@@ -152,6 +225,26 @@ bool VPCodecConfigurationRecord::ParseWebM(const std::vector<uint8_t>& data) {
   return true;
 }
 
+void VPCodecConfigurationRecord::SetVP9Level(uint16_t width,
+                                             uint16_t height,
+                                             double sample_duration_seconds) {
+  // https://www.webmproject.org/vp9/levels/.
+
+  const uint32_t luma_picture_size = width * height;
+  // Alt-Ref frames are not taken into consideration intentionally to avoid the
+  // extra complexities. It may result in smaller luma_sample_rate may than the
+  // actual luma_sample_rate, leading to incorrect level being returned.
+  // If this is a problem, please file a bug to
+  // https://github.com/google/shaka-packager/issues.
+  const double kUnknownSampleDuration = 0.0;
+  // The decision is based on luma_picture_size only if duration is unknown.
+  uint64_t luma_sample_rate = 0;
+  if (sample_duration_seconds != kUnknownSampleDuration)
+    luma_sample_rate = luma_picture_size / sample_duration_seconds;
+
+  level_ = LevelFromCharacteristics(luma_sample_rate, luma_picture_size);
+}
+
 void VPCodecConfigurationRecord::WriteMP4(std::vector<uint8_t>* data) const {
   BufferWriter writer;
   writer.AppendInt(profile());

diff --git a/packager/media/codecs/vp_codec_configuration_record.h b/packager/media/codecs/vp_codec_configuration_record.h
@@ -187,6 +187,11 @@ class VPCodecConfigurationRecord {
   /// @return false if there is parsing errors.
   bool ParseWebM(const std::vector<uint8_t>& data);
 
+  /// Compute and set VP9 Level based on the input attributes.
+  void SetVP9Level(uint16_t width,
+                   uint16_t height,
+                   double sample_duration_seconds);
+
   /// @param data should not be null.
   /// Writes VP codec configuration record to buffer using MP4 format.
   void WriteMP4(std::vector<uint8_t>* data) const;

diff --git a/packager/media/codecs/vp_codec_configuration_record_unittest.cc b/packager/media/codecs/vp_codec_configuration_record_unittest.cc
@@ -161,5 +161,57 @@ TEST(VPCodecConfigurationRecordTest, MergeChromaSubsampling) {
   EXPECT_EQ(AVCHROMA_LOC_TOPLEFT, vp_config.chroma_location());
 }
 
+TEST(VPCodecConfigurationRecordTest, SetLevel) {
+  const uint8_t kUnknownLevel = 0;
+  VPCodecConfigurationRecord vp_config(0x02, kUnknownLevel, 0x08, 0x02, true,
+                                       0x03, 0x04, 0x05,
+                                       std::vector<uint8_t>());
+  ASSERT_EQ(kUnknownLevel, vp_config.level());
+
+  // kExamples are copied from https://www.webmproject.org/vp9/levels/.
+  struct {
+    int expected_level;
+    int width;
+    int height;
+    int frame_rate;
+  } kExamples[] = {
+      {10, 256, 144, 15},   {11, 384, 192, 30},    {20, 480, 256, 30},
+      {21, 640, 384, 30},   {30, 1080, 512, 30},   {31, 1280, 768, 30},
+      {40, 2048, 1088, 30}, {41, 2048, 1088, 60},  {50, 4096, 2176, 30},
+      {51, 4096, 2176, 60}, {52, 4096, 2176, 120}, {60, 8192, 4352, 30},
+      {61, 8192, 4352, 60}, {62, 8192, 4352, 120},
+  };
+  for (const auto& example : kExamples) {
+    vp_config.SetVP9Level(example.width, example.height,
+                          1.0 / example.frame_rate);
+    ASSERT_EQ(example.expected_level, vp_config.level());
+  }
+}
+
+TEST(VPCodecConfigurationRecordTest, SetLevelWithUnknownFrameDuration) {
+  const uint8_t kUnknownLevel = 0;
+  VPCodecConfigurationRecord vp_config(0x02, kUnknownLevel, 0x08, 0x02, true,
+                                       0x03, 0x04, 0x05,
+                                       std::vector<uint8_t>());
+  ASSERT_EQ(kUnknownLevel, vp_config.level());
+
+  // kExamples are modified from https://www.webmproject.org/vp9/levels/ with
+  // frame rate removed.
+  struct {
+    int expected_level;
+    int width;
+    int height;
+  } kExamples[] = {
+      {10, 256, 144},   {11, 384, 192},   {20, 480, 256},
+      {21, 640, 384},   {30, 1080, 512},  {31, 1280, 768},
+      {40, 2048, 1088}, {50, 4096, 2176}, {60, 8192, 4352},
+  };
+  for (const auto& example : kExamples) {
+    const int kUnknownFrameDuration = 0;
+    vp_config.SetVP9Level(example.width, example.height, kUnknownFrameDuration);
+    ASSERT_EQ(example.expected_level, vp_config.level());
+  }
+}
+
 }  // namespace media
 }  // namespace shaka
diff --git a/packager/media/formats/mp4/mp4_media_parser.cc b/packager/media/formats/mp4/mp4_media_parser.cc
@@ -504,6 +504,8 @@ bool MP4MediaParser::ParseMoov(BoxReader* reader) {
       if (desc_idx >= samp_descr.video_entries.size())
         desc_idx = 0;
       const VideoSampleEntry& entry = samp_descr.video_entries[desc_idx];
+      std::vector<uint8_t> codec_configuration_data =
+          entry.codec_configuration.data;
 
       uint32_t coded_width = entry.width;
       uint32_t coded_height = entry.height;
@@ -521,7 +523,7 @@ bool MP4MediaParser::ParseMoov(BoxReader* reader) {
       switch (actual_format) {
         case FOURCC_av01: {
           AV1CodecConfigurationRecord av1_config;
-          if (!av1_config.Parse(entry.codec_configuration.data)) {
+          if (!av1_config.Parse(codec_configuration_data)) {
             LOG(ERROR) << "Failed to parse av1c.";
             return false;
           }
@@ -531,7 +533,7 @@ bool MP4MediaParser::ParseMoov(BoxReader* reader) {
         case FOURCC_avc1:
         case FOURCC_avc3: {
           AVCDecoderConfigurationRecord avc_config;
-          if (!avc_config.Parse(entry.codec_configuration.data)) {
+          if (!avc_config.Parse(codec_configuration_data)) {
             LOG(ERROR) << "Failed to parse avcc.";
             return false;
           }
@@ -568,7 +570,7 @@ bool MP4MediaParser::ParseMoov(BoxReader* reader) {
         case FOURCC_hev1:
         case FOURCC_hvc1: {
           HEVCDecoderConfigurationRecord hevc_config;
-          if (!hevc_config.Parse(entry.codec_configuration.data)) {
+          if (!hevc_config.Parse(codec_configuration_data)) {
             LOG(ERROR) << "Failed to parse hevc.";
             return false;
           }
@@ -579,10 +581,17 @@ bool MP4MediaParser::ParseMoov(BoxReader* reader) {
         case FOURCC_vp08:
         case FOURCC_vp09: {
           VPCodecConfigurationRecord vp_config;
-          if (!vp_config.ParseMP4(entry.codec_configuration.data)) {
+          if (!vp_config.ParseMP4(codec_configuration_data)) {
             LOG(ERROR) << "Failed to parse vpcc.";
             return false;
           }
+          if (actual_format == FOURCC_vp09 &&
+              (!vp_config.is_level_set() || vp_config.level() == 0)) {
+            const double kUnknownSampleDuration = 0.0;
+            vp_config.SetVP9Level(coded_width, coded_height,
+                                  kUnknownSampleDuration);
+            vp_config.WriteMP4(&codec_configuration_data);
+          }
           codec_string = vp_config.GetCodecString(video_codec);
           break;
         }
@@ -606,9 +615,8 @@ bool MP4MediaParser::ParseMoov(BoxReader* reader) {
       std::shared_ptr<VideoStreamInfo> video_stream_info(new VideoStreamInfo(
           track->header.track_id, timescale, duration, video_codec,
           GetH26xStreamFormat(actual_format), codec_string,
-          entry.codec_configuration.data.data(),
-          entry.codec_configuration.data.size(), coded_width, coded_height,
-          pixel_width, pixel_height,
+          codec_configuration_data.data(), codec_configuration_data.size(),
+          coded_width, coded_height, pixel_width, pixel_height,
           0,  // trick_play_factor
           nalu_length_size, track->media.header.language.code, is_encrypted));
 

diff --git a/packager/media/formats/webm/webm_tracks_parser.cc b/packager/media/formats/webm/webm_tracks_parser.cc
@@ -223,16 +223,24 @@ bool WebMTracksParser::OnListEnd(int id) {
         }
         video_default_duration_ = default_duration_;
 
-        // |vp_config_| is only useful for VP8 and VP9.
-        if (codec_id_ == "V_VP8" || codec_id_ == "V_VP9")
-          vp_config_ = video_client_.GetVpCodecConfig(codec_private_);
-
         DCHECK(!video_stream_info_);
         video_stream_info_ = video_client_.GetVideoStreamInfo(
             video_track_num_, codec_id_, codec_private_,
             !video_encryption_key_id_.empty());
         if (!video_stream_info_)
           return false;
+
+        if (codec_id_ == "V_VP8" || codec_id_ == "V_VP9") {
+          vp_config_ = video_client_.GetVpCodecConfig(codec_private_);
+          const double kNanosecondsPerSecond = 1000000000.0;
+          if (codec_id_ == "V_VP9" &&
+              (!vp_config_.is_level_set() || vp_config_.level() == 0)) {
+            vp_config_.SetVP9Level(
+                video_stream_info_->width(), video_stream_info_->height(),
+                video_default_duration_ / kNanosecondsPerSecond);
+          }
+        }
+
       } else {
         DLOG(INFO) << "Ignoring video track " << track_num_;
         ignored_tracks_.insert(track_num_);