Add test scripts for Razel (c1eb02d6) · Commits · IVAS Codec Public Collaboration / IVAS Codec

scripts/razel/ivas_files_map_for_razel.py

0 → 100644

+82 −0

Original line number	Diff line number	Diff line
		from typing import Optional

		from ivas_razel_runner.include.core import (
		AudioConfig,
		AudioFormat,
		AudioSampleType,
		CustomSpeakerLayoutFile,
		File,
		Files,
		SceneDescriptionFile,
		WavFile,
		)
		from ivas_razel_runner.include.ivas_rend import IvasRendInputFile


		def in_file_for_audio_config(config: AudioConfig \| SceneDescriptionFile) -> IvasRendInputFile:
		if config == AudioFormat.MONO:
		return WavFile("scripts/testv/stv48c.wav", True, 48000, 1, AudioSampleType.INT16)
		if config == AudioFormat.STEREO:
		return WavFile("scripts/testv/stvST48c.wav", True, 48000, 2, AudioSampleType.INT16)
		if config == AudioFormat.ISM1:
		return WavFile("scripts/testv/stv1ISM48s.wav", True, 48000, 1, AudioSampleType.INT16)
		if config == AudioFormat.ISM2:
		return WavFile("scripts/testv/stv2ISM48s.wav", True, 48000, 2, AudioSampleType.INT16)
		if config == AudioFormat.ISM3:
		return WavFile("scripts/testv/stv3ISM48s.wav", True, 48000, 3, AudioSampleType.INT16)
		if config == AudioFormat.ISM4:
		return WavFile("scripts/testv/stv4ISM48s.wav", True, 48000, 4, AudioSampleType.INT16)
		if config == AudioFormat.MC_5_1:
		return WavFile("scripts/testv/stv51MC48c.wav", True, 48000, 6, AudioSampleType.INT16)
		if config == AudioFormat.MC_7_1:
		return WavFile("scripts/testv/stv71MC48c.wav", True, 48000, 8, AudioSampleType.INT16)
		if config == AudioFormat.MC_5_1_2:
		return WavFile("scripts/testv/stv512MC48c.wav", True, 48000, 8, AudioSampleType.INT16)
		if config == AudioFormat.MC_5_1_4:
		return WavFile("scripts/testv/stv514MC48c.wav", True, 48000, 10, AudioSampleType.INT16)
		if config == AudioFormat.MC_7_1_4:
		return WavFile("scripts/testv/stv714MC48c.wav", True, 48000, 12, AudioSampleType.INT16)
		if config == AudioFormat.FOA:
		return WavFile("scripts/testv/stvFOA48c.wav", True, 48000, 4, AudioSampleType.INT16)
		if config == AudioFormat.HOA2:
		return WavFile("scripts/testv/stv2OA48c.wav", True, 48000, 9, AudioSampleType.INT16)
		if config == AudioFormat.HOA3:
		return WavFile("scripts/testv/stv3OA48c.wav", True, 48000, 16, AudioSampleType.INT16)
		if config == AudioFormat.MASA1:
		return WavFile("scripts/testv/stv2MASA1TC48c.wav", True, 48000, 1, AudioSampleType.INT16)
		if config == AudioFormat.MASA2:
		return WavFile("scripts/testv/stv2MASA2TC48c.wav", True, 48000, 2, AudioSampleType.INT16)
		if isinstance(config, SceneDescriptionFile):
		return config
		if isinstance(config, CustomSpeakerLayoutFile):
		# Assume test file is 5_1 for now
		assert config.num_channels == AudioFormat.MC_5_1.num_channels
		return WavFile("scripts/testv/stv51MC48c.wav", True, 48000, 6, AudioSampleType.INT16)

		raise RuntimeError(f"No input file for audio config {config}")


		def md_files_for_audio_config(config: AudioConfig \| SceneDescriptionFile) -> Optional[Files]:
		if config == AudioFormat.ISM1:
		return [File("scripts/testv/stvISM1.csv", True)]
		if config == AudioFormat.ISM2:
		return [File("scripts/testv/stvISM1.csv", True), File("scripts/testv/stvISM2.csv", True)]
		if config == AudioFormat.ISM3:
		return [
		File("scripts/testv/stvISM1.csv", True),
		File("scripts/testv/stvISM2.csv", True),
		File("scripts/testv/stvISM3.csv", True),
		]
		if config == AudioFormat.ISM4:
		return [
		File("scripts/testv/stvISM1.csv", True),
		File("scripts/testv/stvISM2.csv", True),
		File("scripts/testv/stvISM3.csv", True),
		File("scripts/testv/stvISM4.csv", True),
		]
		if config == AudioFormat.MASA1:
		return [File("scripts/testv/stv2MASA1TC48c.met", True)]
		if config == AudioFormat.MASA2:
		return [File("scripts/testv/stv2MASA2TC48c.met", True)]

		return None

scripts/razel/test_renderer_razel.py

0 → 100755

+177 −0

Original line number	Diff line number	Diff line
		#!/usr/bin/env python3

		import itertools
		import logging
		from pathlib import Path

		from ivas_files_map_for_razel import in_file_for_audio_config, md_files_for_audio_config
		from ivas_razel_runner.include.core import (
		AudioFile,
		AudioFormat,
		CustomSpeakerLayoutFile,
		File,
		FileFormat,
		SceneDescriptionFile,
		WavDiffArgs,
		WavFile,
		)
		from ivas_razel_runner.include.ivas_rend import IvasRend, IvasRendConfig, UnsupportedConfigWarning
		from ivas_razel_runner.include.sox import Sox
		from razel import Razel

		#######################################################

		BINAURAL_FORMATS = [
		AudioFormat.BINAURAL,
		AudioFormat.BINAURAL_ROOM_IR,
		AudioFormat.BINAURAL_ROOM_REVERB,
		]
		DEFAULT_CUSTOM_SPEAKER_LAYOUT = CustomSpeakerLayoutFile("scripts/ls_layouts/cicp6.txt", 6)
		DEFAULT_SCENE_DESCRIPTION_FILE = SceneDescriptionFile("_dev/in/scene_ism2.txt", in_file_for_audio_config(AudioFormat.ISM2)) # type: ignore

		#######################################################

		INPUT_FORMATS = IvasRend.get_valid_in_audio_configs(DEFAULT_CUSTOM_SPEAKER_LAYOUT, DEFAULT_SCENE_DESCRIPTION_FILE)
		OUTPUT_FORMATS = IvasRend.get_valid_out_audio_configs(DEFAULT_CUSTOM_SPEAKER_LAYOUT)
		HEAD_ROTATION_FILES = [File("scripts/testv/headrot_case00_3000_q.csv", True), None]
		SAMPLING_RATES = [48000] # TODO(sgi): use others

		#######################################################

		KNOWN_BROKEN_BASELINE_CONFIGS = [
		lambda config: config.in_audio_config == AudioFormat.MASA1,
		lambda config: isinstance(config.in_audio_config, SceneDescriptionFile),
		lambda config: config.in_audio_config == AudioFormat.MASA2
		and isinstance(config.out_audio_config, CustomSpeakerLayoutFile),
		lambda config: config.in_audio_config == AudioFormat.MASA2
		and config.out_audio_config in [AudioFormat.BINAURAL_ROOM_IR, AudioFormat.BINAURAL_ROOM_REVERB],
		]
		KNOWN_CLIPPING_CONFIGS = [
		lambda config: config.in_audio_config
		in [AudioFormat.MC_5_1_2, AudioFormat.MC_5_1_4, AudioFormat.MC_5_1, AudioFormat.MC_7_1_4, AudioFormat.MC_7_1]
		and config.out_audio_config in [AudioFormat.MONO, AudioFormat.FOA, AudioFormat.HOA2, AudioFormat.HOA3],
		lambda config: isinstance(config.in_audio_config, CustomSpeakerLayoutFile)
		and config.out_audio_config
		in [AudioFormat.MONO, AudioFormat.STEREO, AudioFormat.FOA, AudioFormat.HOA2, AudioFormat.HOA3],
		lambda config: config.in_audio_config
		in [
		AudioFormat.MC_5_1,
		AudioFormat.MC_5_1_2,
		AudioFormat.MC_5_1_4,
		AudioFormat.MC_7_1,
		AudioFormat.MC_7_1_4,
		AudioFormat.FOA,
		AudioFormat.HOA2,
		AudioFormat.HOA3,
		]
		and config.out_audio_config in BINAURAL_FORMATS,
		]
		NON_BE_TO_REF_WITH_5MS_FRAMING = [
		lambda config: config.in_audio_config
		in [
		AudioFormat.ISM1,
		AudioFormat.ISM2,
		AudioFormat.ISM3,
		AudioFormat.ISM4,
		]
		and config.out_audio_config != AudioFormat.MONO,
		lambda config: config.in_audio_config
		in [
		AudioFormat.MASA1,
		AudioFormat.MASA2,
		],
		]
		NON_BE_TO_REF_WITH_20MS_FRAMING = [
		lambda config: config.out_audio_config in BINAURAL_FORMATS and config.head_rotation is not None,
		]

		#######################################################


		def main():
		root_workspace = Path(__file__).parent / ".." / ".."
		Razel.init(str(root_workspace))
		logging.basicConfig(level=logging.INFO)

		rend_cut = IvasRend(str(root_workspace / "IVAS_rend"))
		rend_ref = IvasRend(str(root_workspace / "IVAS_rend_ref_5ms"))
		sox = Sox("sox")
		AudioFile.set_audio_conversion_tool(sox)

		# Define exceptional configs - order matters here!
		rend_ref.add_support_check_for_nonstandard_configs(match_broken_config)
		rend_cut.add_support_check_for_nonstandard_configs(match_broken_config)
		rend_ref.add_support_check_for_nonstandard_configs(
		# Reference can do binaural rendering with 20 ms frames
		lambda config: config.out_audio_config in BINAURAL_FORMATS
		and config.in_audio_config not in [AudioFormat.MONO, AudioFormat.STEREO]
		and not config.framing_5ms
		)

		for in_format, out_format, head_rotation_file in itertools.product(
		INPUT_FORMATS, OUTPUT_FORMATS, HEAD_ROTATION_FILES
		):
		in_file = in_file_for_audio_config(in_format)
		md_files = md_files_for_audio_config(in_format)
		config = IvasRendConfig(in_format, out_format, md_files, head_rotation=head_rotation_file, framing_5ms=False)

		if any(match(config) for match in KNOWN_CLIPPING_CONFIGS):
		config.input_gain = 0.2

		# These files' length is not a multiple of 5ms, which causes different output lengths with 5ms framing on/off
		if in_file.basename in ["stvFOA48c.wav", "stv2OA48c.wav", "stv3OA48c.wav"]:
		assert isinstance(in_file, WavFile)
		in_file = in_file.trim(0, 20)

		# Run REF 20ms
		try:
		out_20ms_ref = rend_ref.render(in_file, config, FileFormat.WAV)
		except UnsupportedConfigWarning as e:
		logging.info("%s Skipping %s", e.reason, config)
		out_20ms_ref = None

		# Run CUT 20ms
		try:
		out_20ms_cut = rend_cut.render(in_file, config, FileFormat.WAV)
		except UnsupportedConfigWarning as e:
		logging.info("%s Skipping %s", e.reason, config)
		out_20ms_cut = None

		if out_20ms_ref and out_20ms_cut:
		if any(match(config) for match in NON_BE_TO_REF_WITH_20MS_FRAMING):
		assert isinstance(out_20ms_cut, WavFile)
		out_20ms_cut.should_not_equal(out_20ms_ref)
		out_20ms_cut.should_be_similar_to(out_20ms_ref, WavDiffArgs(mld_max=0.01))
		else:
		out_20ms_cut.should_equal(out_20ms_ref)

		# Run CUT 5ms
		config.framing_5ms = True
		try:
		out_5ms_cut = rend_cut.render(in_file, config, FileFormat.WAV)
		except UnsupportedConfigWarning as e:
		logging.info("%s Skipping %s", e.reason, config)
		out_5ms_cut = None

		if out_5ms_cut and out_20ms_ref:
		if any(match(config) for match in NON_BE_TO_REF_WITH_5MS_FRAMING):
		assert isinstance(out_5ms_cut, WavFile)
		out_5ms_cut.should_not_equal(out_20ms_ref)
		out_5ms_cut.should_be_similar_to(
		out_20ms_ref, WavDiffArgs(mld_max=16)
		) # TODO(sgi): also test ISM with no metadata (should be BE)
		else:
		out_5ms_cut.should_equal(out_20ms_ref)

		Razel.instance().run(["exec", "-k"])


		def match_broken_config(config):
		if any(match(config) for match in KNOWN_BROKEN_BASELINE_CONFIGS):
		raise UnsupportedConfigWarning("Known broken config.")

		return False


		if __name__ == "__main__":
		main()