Commit 3f45c63a authored by Archit Tamarapu's avatar Archit Tamarapu
Browse files

Merge branch 'fix_osba_rendering_scaling' into 'main' 

[fix] scaling of ISM and SBA rendered output before combining for OSBA

See merge request !156
parents ca9a4d98 8256128f

ivas_processing_scripts/audiotools/__init__.py

+8 −4
Original line number Diff line number Diff line
@@ -244,7 +244,9 @@ def main(): 
    # validate required arguments

    if args.input is None or args.in_fmt is None or args.output is None:

        parser.print_usage()

        raise SystemExit("the following arguments are required: -i/--in, -if/--in_fmt, -o/--out")

        raise SystemExit(

            "the following arguments are required: -i/--in, -if/--in_fmt, -o/--out"

        )



    elif args.input is not None:

        if not args.out_fs:
@@ -280,7 +282,9 @@ def main(): 
        else:

            # input was a dir so output should be a dir too

            if args.output.is_file():

                raise NotADirectoryError(f"Input directory '{args.input}' specified with file output to '{args.output}' - please specify an output directory instead")

                raise NotADirectoryError(

                    f"Input directory '{args.input}' specified with file output to '{args.output}' - please specify an output directory instead"

                )

            if not args.output.exists():

                args.output.mkdir()

                print(f"Created output directory {args.output}")

ivas_processing_scripts/audiotools/convert/osba.py

+10 −9
Original line number Diff line number Diff line
@@ -81,14 +81,14 @@ def convert_osba( 


        # combine results

        try:

            out.audio = out_ism.audio + out_sba.audio

            out.audio = 0.5 * (out_ism.audio + out_sba.audio)

        except ValueError as e:

            if (

                out_ism.audio.shape[0] != out_sba.audio.shape[0]

                and out_sba.audio.shape[0] == osba.audio.shape[0]

            ):

                out.audio = out_sba.audio

                out.audio[: out_ism.audio.shape[0], :] += out_ism.audio

                out.audio = 0.5 * out_sba.audio

                out.audio[: out_ism.audio.shape[0], :] += 0.5 * out_ism.audio

            else:

                raise e
@@ -104,7 +104,7 @@ def convert_osba( 
        render_sba_to_cba(sba, out_sba)



        # combine results

        out.audio = out_ism.audio + out_sba.audio

        out.audio = 0.5 * (out_ism.audio + out_sba.audio)



    # OSBA -> SBA

    elif isinstance(out, audio.SceneBasedAudio):
@@ -118,7 +118,7 @@ def convert_osba( 
        render_sba_to_sba(sba, out_sba)



        # combine results

        out.audio = out_ism.audio + out_sba.audio

        out.audio = 0.5 * (out_ism.audio + out_sba.audio)



    # OSBA -> OMASA

    elif isinstance(out, audio.OMASAAudio):
@@ -128,14 +128,15 @@ def convert_osba( 
                "OSBA to OMASA conversion only possible if number of ISM objects matches"

            )



        # only render SBA part

        # only render SBA part to MASA

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

        out_sba.metadata_file = out.metadata_files[-1]

        render_sba_to_masa(sba, out_sba)



        out.audio = np.concatenate(

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

        )

        # ism audio is passed through

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



        out.audio = np.concatenate((ism_audio, out_sba.audio), axis=1)



    # OSBA -> OSBA

    elif isinstance(out, audio.OSBAAudio):

ivas_processing_scripts/audiotools/convert/scenebased.py

+3 −3
Original line number Diff line number Diff line
@@ -35,7 +35,7 @@ from typing import Optional, Union 
from warnings import warn



import numpy as np

from scipy.special import lpmv

from scipy.special import factorial, lpmv



from ivas_processing_scripts.audiotools import audio

from ivas_processing_scripts.audiotools.audioarray import delay, framewise_io
@@ -348,8 +348,8 @@ def getRSH( 
    for i, (l, m) in enumerate(LM):

        # N3D norm

        response[i, :] = np.sqrt(

            ((2 * l + 1) * float(np.math.factorial(l - np.abs(m))))

            / (4 * np.pi * float(np.math.factorial(l + np.abs(m))))

            ((2 * l + 1) * float(factorial(l - np.abs(m))))

            / (4 * np.pi * float(factorial(l + np.abs(m))))

        )



        # trig term