Commit 73a29eb9 authored by Vladimir Malenovsky's avatar Vladimir Malenovsky
Browse files

fix the duration parameter - only trim if duration is exceeded

parent 7c150828

ivas_processing_scripts/generation/generate_ismN_items.py

+7 −11
Original line number Diff line number Diff line
@@ -443,18 +443,14 @@ def generate_ismN_scene( 
        noise = np.random.randint(low=-4, high=5, size=y.audio.shape).astype("float")

        y.audio += noise



    # adjust the length of the output signal

    # trim the output signal such if the total duration exceeds X seconds

    if "duration" in cfg.__dict__:

        # trim the output signal such that the total duration is X seconds

        duration = int(cfg.duration * cfg.fs)  # convert to samples

    else:

        # do not change the length of the audio signal

        duration = len(y.audio)

    duration = int(

        np.floor(duration / frame_len) * frame_len

    )  # ensure multiple of 20ms

    if len(y.audio) != duration:

        metadata.trim_meta(y, limits=[0, len(y.audio) - duration], samples=True)

        # convert from seconds to samples (ensure multiple of 20ms)

        duration = int(np.floor(int(cfg.duration * cfg.fs) / frame_len) * frame_len)



        # check if the current length of the output signal exceeds the duration

        if len(y.audio) > duration:

            metadata.trim_meta(y, limits=[0, duration], samples=True)



    # adjust the loudness of the output signal

    if "loudness" in cfg.__dict__:

ivas_processing_scripts/generation/generate_masa_items.py

+14 −16
Original line number Diff line number Diff line
@@ -316,12 +316,12 @@ def generate_MASA_scene( 
            x = reverb_hoa3(x, IR, mode=None)



        # adjust the level of the FOA/HOA2/HOA3 signal

        if level is None:

            # do not change the level of the audio source signal

            logger.info("-- Level of the audio source signal is not changed")

        elif np.isinf(level):

        if np.isinf(level):

            # set all channels to zero

            x.audio = np.zeros_like(x.audio)

        elif level is None:

            # do not change the level of the audio source signal

            logger.info("-- Level of the audio source signal is not changed")

        else:

            x.audio, _ = loudness_norm(x, level, loudness_format="STEREO")
@@ -393,17 +393,15 @@ def generate_MASA_scene( 
                y_int.audio, y_int.fs, limits=[-preamble, -postamble], samples=True

            )



    # adjust the length of the output signal

    # trim the output signal if the total duration exceeds X seconds

    if "duration" in cfg.__dict__:

        # trim the output signal such that the total duration is X seconds

        duration = int(cfg.duration * cfg.fs)  # convert to samples

    else:

        # do not change the length of the audio signal

        duration = len(y_int.audio)

        # convert from seconds to samples (ensure multiple of 20ms)

        duration = int(

        np.floor(duration / frame_len) * frame_len

    )  # ensure multiple of 20ms

    if len(y_int.audio) != duration:

            np.floor(int(cfg.duration * cfg.fs) / frame_len) * frame_len

        )



        # check if the current length of the output signal exceeds the duration

        if len(y_int.audio) > duration:

            y_int.audio = audioarray.trim(

                y_int.audio, y_int.fs, limits=[0, len(y_int.audio) - duration], samples=True

            )

ivas_processing_scripts/generation/generate_mc_items.py

+10 −12
Original line number Diff line number Diff line
@@ -393,17 +393,15 @@ def generate_MC_scene( 
                y_int.audio, y_int.fs, limits=[-preamble, -postamble], samples=True

            )



    # adjust the length of the output signal

    # trim the output signal if the total duration exceeds X seconds

    if "duration" in cfg.__dict__:

        # trim the output signal such that the total duration is X seconds

        duration = int(cfg.duration * cfg.fs)  # convert to samples

    else:

        # do not change the length of the audio signal

        duration = len(y_int.audio)

        # convert from seconds to samples (ensure multiple of 20ms)

        duration = int(

        np.floor(duration / frame_len) * frame_len

    )  # ensure multiple of 20ms

    if len(y_int.audio) != duration:

            np.floor(int(cfg.duration * cfg.fs) / frame_len) * frame_len

        )



        # check if the current length of the output signal exceeds the duration

        if len(y_int.audio) > duration:

            y_int.audio = audioarray.trim(

                y_int.audio, y_int.fs, limits=[0, len(y_int.audio) - duration], samples=True

            )

ivas_processing_scripts/generation/generate_omasa_items.py

+12 −1
Original line number Diff line number Diff line
@@ -472,8 +472,9 @@ def generate_OMASA_scene( 
        noise = np.random.randint(low=-4, high=5, size=y.audio.shape).astype("float")

        y_int.audio += noise



    # adjust the length of the output signal

    # trim the output signal such if the total duration exceeds X seconds

    if "duration" in cfg.__dict__:

<<<<<<< Updated upstream

        # trim the output signal such that the total duration is X seconds

        duration = int(cfg.duration * cfg.fs)  # convert to samples

    else:
@@ -484,6 +485,16 @@ def generate_OMASA_scene( 
    )  # ensure multiple of 20ms

    if len(y_int.audio) != duration:

        metadata.trim_meta(y_int, limits=[0, len(y_int.audio) - duration], samples=True)

=======

        # convert from seconds to samples (ensure multiple of 20ms)

        duration = int(

            np.floor(int(cfg.duration * cfg.fs) / frame_len) * frame_len

        )



        # check if the current length of the output signal exceeds the duration

        if len(y.audio) > duration:

            metadata.trim_meta(y, limits=[0, len(y.audio) - duration], samples=True)

>>>>>>> Stashed changes



    # adjust the loudness of the output signal

    if "loudness" in cfg.__dict__:

ivas_processing_scripts/generation/generate_osba_items.py

+9 −11
Original line number Diff line number Diff line
@@ -458,17 +458,15 @@ def generate_OSBA_scene( 
        noise = np.random.randint(low=-4, high=5, size=y.audio.shape).astype("float")

        y.audio += noise



    # adjust the length of the output signal

    # trim the output signal if the total duration exceeds X seconds

    if "duration" in cfg.__dict__:

        # trim the output signal such that the total duration is X seconds

        duration = int(cfg.duration * cfg.fs)  # convert to samples

    else:

        # do not change the length of the audio signal

        duration = len(y.audio)

        # convert from seconds to samples (ensure multiple of 20ms)

        duration = int(

        np.floor(duration / frame_len) * frame_len

    )  # ensure multiple of 20ms

    if len(y.audio) != duration:

            np.floor(int(cfg.duration * cfg.fs) / frame_len) * frame_len

        )



        # check if the current length of the output signal exceeds the duration

        if len(y.audio) > duration:

            metadata.trim_meta(y, limits=[0, len(y.audio) - duration], samples=True)



    # adjust the loudness of the output signal