Commit b8de52cb authored by Anika Treffehn's avatar Anika Treffehn
Browse files

enabled osba to omasa conversion

parent 07ee6f4d

ivas_processing_scripts/audiotools/audio.py

+10 −6
Original line number Diff line number Diff line
@@ -218,18 +218,18 @@ class MetadataAssistedSpatialAudio(Audio): 
            raise ValueError(

                f"Unsupported metadata assisted spatial audio format {name}"

            )

        self.metadata_files = []

        self.metadata_file = None



    @classmethod

    def _from_file(

        cls,

        name: str,

        filename: Path,

        metadata_files: list[str],

        metadata_file: str,

        fs: Optional[int] = None,

    ) -> "MetadataAssistedSpatialAudio":

        obj = super()._from_file(name, filename, fs)

        obj.metadata_file = Path(metadata_files[0])

        obj.metadata_file = Path(metadata_file)

        return obj



    @classmethod
@@ -237,11 +237,11 @@ class MetadataAssistedSpatialAudio(Audio): 
        cls,

        name: str,

        filename: Path,

        metadata_files: list[str],

        metadata_file: str,

        fs: Optional[int] = None,

    ) -> "MetadataAssistedSpatialAudio":

        obj = super()._from_file(name, filename, fs)

        obj.metadata_file = Path(metadata_files[0])

        obj.metadata_file = Path(metadata_file)

        return obj
@@ -432,13 +432,17 @@ class OMASAAudio(Audio): 


    def init_metadata(self):

        # check if number of metadata files matches format

        if self.num_ism_channels != len(self.metadata_files):

        if self.num_ism_channels != len(self.metadata_files)-1:

            raise ValueError(

                f"Mismatch between number of ism channels [{self.num_ism_channels}], and metadata [{len(self.metadata_files)}]"

            )



        self.object_pos = []

        for i, f in enumerate(self.metadata_files):

            if i >= self.num_ism_channels:

                # only read ISM metadata, not MASA metadata

                break



            pos = np.genfromtxt(f, delimiter=",")



            # check if metadata has right number of columns

ivas_processing_scripts/audiotools/convert/__init__.py

+42 −13
Original line number Diff line number Diff line
@@ -34,6 +34,7 @@ import logging 
from pathlib import Path, PurePath

from shutil import copyfile

from typing import Optional, Union

from copy import copy



from numpy import empty
@@ -79,9 +80,13 @@ def convert_file( 
    else:

        # first check prevents crash on custom_ls setup formats

        if isinstance(in_fmt, str) and in_fmt.startswith("MASA") and in_meta is None:

            # TODO treffehn: also consider OMASA

            # if no MD file is provided, default to name (including .wav or .pcm!!!) + ".met"

            in_meta = [in_file.parent / (in_file.name + ".met")]

        elif isinstance(in_fmt, str) and "MASA" in in_fmt and "ISM" in in_fmt:

            if isinstance(in_meta, list) and len(in_meta) < (int(in_fmt[3])+int(in_fmt[8])):

                in_meta_masa = in_file.parent / (in_file.name + ".met")

                in_meta.append(in_meta_masa)



        input = audio.fromfile(in_fmt, in_file, in_fs, in_meta)



    # try to set reasonable defaults if missing
@@ -100,22 +105,40 @@ def convert_file( 


    output = audio.fromtype(out_fmt)



    # handle metadata for outputs with metadata (MASA, ISM, OMASA, OSBA)

    if isinstance(output, audio.MetadataAssistedSpatialAudio):

        # create dummy audio array to allow inference of MASA mode

        num_tcs = int(output.name[-1])

        output.audio = empty((1, num_tcs))



        if isinstance(input, audio.MetadataAssistedSpatialAudio):

            # use existing metadata file

            output.metadata_file = input.metadata_file

        else:

            # fabricate metadata file name

        output.metadata_file = Path(out_file).parent / (Path(out_file).name + ".met")

    if isinstance(output, audio.ObjectBasedAudio):

            masa_meta_file_name = Path(out_file).parent / (Path(out_file).name + ".met")

            output.metadata_file = masa_meta_file_name



    elif isinstance(output, audio.ObjectBasedAudio):

        try:

            output.object_pos = input.object_pos

            output.metadata_files = input.metadata_files

            output.object_pos = copy(input.object_pos)

            output.metadata_files = copy(input.metadata_files)

        except Exception:

            raise ValueError(

                "ISM is not supported as an output for rendering! Only usable as pass-through"

            )

    elif isinstance(output, audio.OMASAAudio):

        if isinstance(input, audio.OMASAAudio):

            # use existing metadata files

            output.metadata_files = copy(input.metadata_files)

        else:

            # fabricate metadata file name

            masa_meta_file_name = Path(out_file).parent / (Path(out_file).name + ".met")

            output.metadata_files = copy(input.metadata_files)

            output.metadata_files.append(masa_meta_file_name)

        output.object_pos = copy(input.object_pos)



    elif isinstance(output, audio.OSBAAudio):

        output.object_pos = copy(input.object_pos)

        output.metadata_files = copy(input.metadata_files)



    # apply actual conversion

    if isinstance(input, metadata.Metadata):

        if logger:

            logger.debug(f"Converting metadata to {out_fmt} : {in_file} -> {out_file}")
@@ -137,16 +160,22 @@ def convert_file( 
        output.fs = in_fs  # resampling not yet applied

        convert(input, output, in_fs=in_fs, out_fs=out_fs, logger=logger, **kwargs)



    # write output

    # write output audio

    write(out_file, output.audio, output.fs)

    if isinstance(output, audio.ObjectBasedAudio):

    # write metadata

    if isinstance(output, audio.ObjectBasedAudio) or isinstance(output, audio.OSBAAudio):

        write_ISM_metadata_in_file(output.object_pos, [out_file], automatic_naming=True)

    elif isinstance(output, audio.MetadataAssistedSpatialAudio) and in_fmt == out_fmt:

        # audio objects point to same MD file, create new one with default naming for output

        out_md_name = out_file.parent / (out_file.name + ".met")

        copyfile(output.metadata_file, out_md_name)

        output.metadata_file = out_md_name



    elif isinstance(output, audio.OMASAAudio):

        write_ISM_metadata_in_file(output.object_pos, [out_file], automatic_naming=True)

        if in_fmt == out_fmt:

            # audio objects point to same MD file, create new one with default naming for output

            out_md_name = out_file.parent / (out_file.name + ".met")

            copyfile(output.metadata_files[-1], out_md_name)



def convert(

    input: audio.Audio,

ivas_processing_scripts/audiotools/convert/osba.py

+14 −12
Original line number Diff line number Diff line
@@ -29,7 +29,8 @@ 
#  the United Nations Convention on Contracts on the International Sales of Goods.

#



import copy

from copy import copy, deepcopy

import numpy as np



from ivas_processing_scripts.audiotools import audio

from ivas_processing_scripts.audiotools.convert.objectbased import render_oba_to_binaural, render_oba_to_cba, \
@@ -49,19 +50,19 @@ def convert_osba( 


    # split OSBA object in ISM and SBA object

    oba = audio.fromarray("ISM" + str(osba.num_ism_channels), osba.audio[:, :osba.num_ism_channels], osba.fs)

    oba.metadata_files = osba.metadata_files

    oba.object_pos = osba.object_pos

    oba.metadata_files = copy(osba.metadata_files)

    oba.object_pos = copy(osba.object_pos)

    sba = audio.fromarray("SBA" + str(osba.ambi_order), osba.audio[:, osba.num_ism_channels:], osba.fs)



    # OSBA -> Binaural

    if isinstance(out, audio.BinauralAudio):

        # render SBA and ISM part separately

        # ISM

        out_ism = copy.deepcopy(out)

        out_ism = deepcopy(out)

        render_oba_to_binaural(oba, out_ism, **kwargs)



        # SBA

        out_sba = copy.deepcopy(out)

        out_sba = deepcopy(out)

        render_sba_to_binaural(sba, out_sba, **kwargs)



        # combine results
@@ -71,11 +72,11 @@ def convert_osba( 
    elif isinstance(out, audio.ChannelBasedAudio):

        # render SBA and ISM part separately

        # ISM

        out_ism = copy.deepcopy(out)

        out_ism = deepcopy(out)

        render_oba_to_cba(oba, out_ism)



        # SBA

        out_sba = copy.deepcopy(out)

        out_sba = deepcopy(out)

        render_sba_to_cba(sba, out_sba)



        # combine results
@@ -85,11 +86,11 @@ def convert_osba( 
    elif isinstance(out, audio.SceneBasedAudio):

        # render SBA and ISM part separately

        # ISM

        out_ism = copy.deepcopy(out)

        out_ism = deepcopy(out)

        render_oba_to_sba(oba, out_ism)



        # SBA

        out_sba = copy.deepcopy(out)

        out_sba = deepcopy(out)

        render_sba_to_sba(sba, out_sba)



        # combine results
@@ -97,17 +98,18 @@ def convert_osba( 


    # OSBA -> OMASA

    elif isinstance(out, audio.OMASAAudio):

        # TODO (treffehn)

        # check if ism object number is the same

        if out.num_ism_channels != osba.num_ism_channels:

            raise ValueError("OSBA to OMASA conversion only possible if number of ISM objects matches")



        # only render SBA part

        out_sba = audio.fromtype(out.name[4:])

        out_sba.metadata_file = out.metadata_files[-1]

        render_sba_to_masa(sba, out_sba)



        out.audio[:, :osba.num_ism_channels] = osba.audio[:, :osba.num_ism_channels]

        out.audio[:, osba.num_ism_channels:] = out_sba.audio

        # out.audio[:, :osba.num_ism_channels] = osba.audio[:, :osba.num_ism_channels]

        # out.audio[:, osba.num_ism_channels:] = out_sba.audio

        out.audio = np.concatenate((osba.audio[:, :osba.num_ism_channels], out_sba.audio), axis=1)



    else:

        raise NotImplementedError(

ivas_processing_scripts/audiotools/convert/scenebased.py

+1 −1
Original line number Diff line number Diff line
@@ -198,7 +198,7 @@ def render_sba_to_masa( 
) -> None:



    num_tcs = masa_out.num_channels

    md_out_path = masa_out.metadata_files  # TODO: get metadata files

    md_out_path = masa_out.metadata_file



    masa = masaAnalyzer(sba_in, num_tcs, masa_out.dirs, md_out_path)

    masa_out.audio = masa.audio

ivas_processing_scripts/audiotools/wrappers/masaAnalyzer.py

+1 −1
Original line number Diff line number Diff line
@@ -106,6 +106,6 @@ def masaAnalyzer( 
        run(cmd, cwd=binary.resolve().parent)



        fmt = f"MASA{num_tcs}DIR{num_dirs}"

        masa = audio.fromfile(fmt, tmp_out_pcm, 48000, [metadata_out_path])

        masa = audio.fromfile(fmt, tmp_out_pcm, 48000, metadata_out_path)



        return masa