Merge pull request #153 from jwcullen/iamf_v2_test_vectors

Upload V2 test vectors

Author: jwcullen

tests/proto/arbitrary_obu.proto

@@ -8,13 +8,17 @@
 // source code in the PATENTS file, you can obtain it at
 // www.aomedia.org/license/patent.
 
-syntax = "proto2";
+edition = "2023";
 
 package libiamf_proto;
 
 import "obu_header.proto";
 
+option features.enum_type = CLOSED;
+
 enum ArbitraryObuType {
+  option allow_alias = true;
+
   OBU_IA_INVALID = 0;
   OBU_IA_CODEC_CONFIG = 1;
   OBU_IA_AUDIO_ELEMENT = 2;
@@ -40,7 +44,8 @@ enum ArbitraryObuType {
   OBU_IA_AUDIO_FRAME_ID_15 = 22;
   OBU_IA_AUDIO_FRAME_ID_16 = 23;
   OBU_IA_AUDIO_FRAME_ID_17 = 24;
-  OBU_IA_RESERVED_24 = 25;
+  OBU_IA_RESERVED_24 = 25 [deprecated = true];
+  OBU_IA_METADATA = 25;
   OBU_IA_RESERVED_25 = 26;
   OBU_IA_RESERVED_26 = 27;
   OBU_IA_RESERVED_27 = 28;
@@ -52,7 +57,11 @@ enum ArbitraryObuType {
 
 enum InsertionHook {
   INSERTION_HOOK_INVALID = 0;
+  // Places this before the canonical descriptors. In an MP4 context, this would
+  // be early in the "IACB" box.
   INSERTION_HOOK_BEFORE_DESCRIPTORS = 1;
+  // Places this after the canonical descriptors. In an MP4 context, this would
+  // be late in the "IACB" box.
   INSERTION_HOOK_AFTER_DESCRIPTORS = 2;
   INSERTION_HOOK_AFTER_IA_SEQUENCE_HEADER = 3;
   INSERTION_HOOK_AFTER_CODEC_CONFIGS = 4;
@@ -64,13 +73,15 @@ enum InsertionHook {
 }
 
 message ArbitraryObuMetadata {
-  optional InsertionHook insertion_hook = 1;
-  optional ArbitraryObuType obu_type = 4;
-  optional ObuHeaderMetadata obu_header = 2;
-  optional bytes payload = 3;
+  InsertionHook insertion_hook = 1;
+  ArbitraryObuType obu_type = 4;
+  ObuHeaderMetadata obu_header = 2;
+  bytes payload = 3;
+
   // Used to force the bitstream to be invalid when this OBU is inserted.
-  optional bool invalidates_bitstream = 5;
+  bool invalidates_bitstream = 5;
+
   // Used to control the temporal unit when using the *_AT_TICK insertion hooks.
   // Ignored otherwise.
-  optional int64 insertion_tick = 6;
+  int64 insertion_tick = 6;
 }

tests/proto/audio_element.proto

@@ -8,32 +8,41 @@
 // source code in the PATENTS file, you can obtain it at
 // www.aomedia.org/license/patent.
 
-syntax = "proto2";
+edition = "2023";
 
 package libiamf_proto;
 
 import "obu_header.proto";
 import "param_definitions.proto";
 
+option features.enum_type = CLOSED;
+
 enum AudioElementType {
   AUDIO_ELEMENT_INVALID = 0;
   AUDIO_ELEMENT_CHANNEL_BASED = 1;
   AUDIO_ELEMENT_SCENE_BASED = 2;
+  AUDIO_ELEMENT_OBJECT_BASED = 3;
 }
 
 message ParamDefinitionExtension {
-  optional uint32 param_definition_size = 1;
-  optional bytes param_definition_bytes = 2;
+  // `param_definition_size` is ignored. The value in the bitstream is inferred
+  // based on the size of `param_definition_bytes`.
+  uint32 param_definition_size = 1 [deprecated = true];
+  bytes param_definition_bytes = 2;
 }
 
 message AudioElementParam {
-  optional uint32 deprecated_param_definition_type = 1 [deprecated = true];
-  optional ParamDefinitionType param_definition_type = 5;
+  uint32 deprecated_param_definition_type = 1 [deprecated = true];
+
+  ParamDefinitionType param_definition_type = 5;
+
   oneof param_definition {
     DemixingParamDefinition demixing_param = 2;
     ReconGainParamDefinition recon_gain_param = 3;
     ParamDefinitionExtension param_definition_extension = 4;
   }
+
+  // Next ID: 6
 }
 
 // Valid proto enums start at index 1, which are different from the
@@ -70,27 +79,36 @@ enum ExpandedLoudspeakerLayout {
   EXPANDED_LOUDSPEAKER_LAYOUT_STEREO_SI = 11;
   EXPANDED_LOUDSPEAKER_LAYOUT_STEREO_TP_SI = 12;
   EXPANDED_LOUDSPEAKER_LAYOUT_TOP_6_CH = 13;
-  EXPANDED_LOUDSPEAKER_LAYOUT_RESERVED_13 = 14;
+  EXPANDED_LOUDSPEAKER_LAYOUT_10_2_9_3_CH = 14;
+  EXPANDED_LOUDSPEAKER_LAYOUT_LFE_PAIR = 15;
+  EXPANDED_LOUDSPEAKER_LAYOUT_BOTTOM_3_CH = 16;
+  EXPANDED_LOUDSPEAKER_LAYOUT_7_1_5_4_CH = 17;
+  EXPANDED_LOUDSPEAKER_LAYOUT_BOTTOM_4_CH = 18;
+  EXPANDED_LOUDSPEAKER_LAYOUT_TOP_1_CH = 19;
+  EXPANDED_LOUDSPEAKER_LAYOUT_TOP_5_CH = 20;
   EXPANDED_LOUDSPEAKER_LAYOUT_RESERVED_255 = 256;
 }
 
 message ChannelAudioLayerConfig {
-  optional uint32 deprecated_loudspeaker_layout = 1 [deprecated = true];
-  optional LoudspeakerLayout loudspeaker_layout = 10;
-  optional uint32 output_gain_is_present_flag = 2;
-  optional uint32 recon_gain_is_present_flag = 3;
-  optional uint32 reserved_a = 4;
-  optional uint32 substream_count = 5;
-  optional uint32 coupled_substream_count = 6;
-  optional uint32 output_gain_flag = 7;
-  optional uint32 reserved_b = 8;
-  optional int32 output_gain = 9;
-  optional ExpandedLoudspeakerLayout expanded_loudspeaker_layout = 11;
+  uint32 deprecated_loudspeaker_layout = 1 [deprecated = true];
+
+  LoudspeakerLayout loudspeaker_layout = 10;
+  uint32 output_gain_is_present_flag = 2;
+  uint32 recon_gain_is_present_flag = 3;
+  uint32 reserved_a = 4;
+  uint32 substream_count = 5;
+  uint32 coupled_substream_count = 6;
+  uint32 output_gain_flag = 7;
+  uint32 reserved_b = 8;
+  int32 output_gain = 9;
+  ExpandedLoudspeakerLayout expanded_loudspeaker_layout = 11;
 }
 
 message ScalableChannelLayoutConfig {
-  optional uint32 num_layers = 1;
-  optional uint32 reserved = 2;
+  // `num_layers` is ignored. The value in the bitstream is inferred based on
+  // the number of `channel_audio_layer_configs`.
+  uint32 num_layers = 1 [deprecated = true];
+  uint32 reserved = 2;
   repeated ChannelAudioLayerConfig channel_audio_layer_configs = 3;
 }
 
@@ -101,44 +119,59 @@ enum AmbisonicsMode {
 }
 
 message AmbisonicsMonoConfig {
-  optional uint32 output_channel_count = 1;
-  optional uint32 substream_count = 2;
-  repeated uint32 channel_mapping = 3 [packed = true];
+  uint32 output_channel_count = 1;
+  uint32 substream_count = 2;
+  repeated uint32 channel_mapping = 3;
 }
 
 message AmbisonicsProjectionConfig {
-  optional uint32 output_channel_count = 1;
-  optional uint32 substream_count = 2;
-  optional uint32 coupled_substream_count = 3;
-  repeated int32 demixing_matrix = 4 [packed = true];
+  uint32 output_channel_count = 1;
+  uint32 substream_count = 2;
+  uint32 coupled_substream_count = 3;
+  repeated int32 demixing_matrix = 4;
 }
 
 message AmbisonicsConfig {
-  optional AmbisonicsMode ambisonics_mode = 1;
+  AmbisonicsMode ambisonics_mode = 1;
+
   oneof ambisonics_config {
     AmbisonicsMonoConfig ambisonics_mono_config = 2;
     AmbisonicsProjectionConfig ambisonics_projection_config = 3;
   }
 }
 
+message ObjectsConfig {
+  uint32 num_objects = 1;
+  bytes objects_config_extension_bytes = 2;
+}
+
 message AudioElementConfigExtension {
-  optional uint32 audio_element_config_size = 1;
-  optional bytes audio_element_config_bytes = 2;
+  // `audio_element_config_size` is ignored. The value in the bitstream is
+  // inferred based on the size of `audio_element_config_bytes`.
+  uint32 audio_element_config_size = 1 [deprecated = true];
+  bytes audio_element_config_bytes = 2;
 }
 
 message AudioElementObuMetadata {
-  optional uint32 audio_element_id = 1;
-  optional AudioElementType audio_element_type = 2;
-  optional uint32 reserved = 3;
-  optional uint32 codec_config_id = 4;
-  optional uint32 num_substreams = 5;
-  repeated uint32 audio_substream_ids = 6 [packed = true];
-  optional uint32 num_parameters = 7;
+  uint32 audio_element_id = 1;
+  AudioElementType audio_element_type = 2;
+  uint32 reserved = 3;
+  uint32 codec_config_id = 4;
+  // `num_substreams` is ignored. The value in the bitstream is inferred based
+  // on the number of `audio_substream_ids`.
+  uint32 num_substreams = 5 [deprecated = true];
+  repeated uint32 audio_substream_ids = 6;
+  // `num_parameters` is ignored. The value in the bitstream is inferred based
+  // on the number of `audio_element_params`.
+  uint32 num_parameters = 7 [deprecated = true];
   repeated AudioElementParam audio_element_params = 8;
+
   oneof config {
     ScalableChannelLayoutConfig scalable_channel_layout_config = 9;
     AmbisonicsConfig ambisonics_config = 10;
     AudioElementConfigExtension audio_element_config_extension = 11;
+    ObjectsConfig objects_config = 13;
   }
-  optional ObuHeaderMetadata obu_header = 12;
+
+  ObuHeaderMetadata obu_header = 12;
 }