Commit 33e28543 authored by Archit Tamarapu's avatar Archit Tamarapu
Browse files

[fix] use correct bandwidth file for reference loudness

parent 5f621f43

scripts/get_loudness_data.py

+41 −7
Original line number Diff line number Diff line
@@ -73,6 +73,7 @@ CFG_TO_PROC_FMT = { 


PATTERN_BITRATE = r"b([\d_]*|all)(_dtx)?_(swb|wb|fb)"

RESULT_OUTPUT_FILE = Path(__file__).parent.parent.joinpath("loudness.csv")

BANDWIDTH_TO_FS = {"wb": "16", "swb": "32", "fb": "48"}





def find_metadata_files(format, file):
@@ -241,15 +242,34 @@ def main(args): 
        output_folder = OUTPUT_FOLDER

        pattern = re.compile(rf"{re.escape(cfg_fmt)}_{PATTERN_BITRATE}")



        # Load and cache input audio once per format

        if proc_fmt not in input_audio_cache:

        # Load and cache input audio for each format and bandwidth combination

        for bw in BANDWIDTH_TO_FS.keys():

            cache_key = (proc_fmt, bw)



            if cache_key not in input_audio_cache:

                fs = BANDWIDTH_TO_FS[bw]

                pcm_folder = OUTPUT_FOLDER.parent.joinpath("pcm")

                resampled_file = (

                    pcm_folder

                    / f"{infile.stem}_{fs}_{replacements.get(cfg_fmt, cfg_fmt)}.pcm"

                )



                if not resampled_file.exists():

                    print(

                        f"\n⚠️  Resampled file not found: {resampled_file}",

                        file=sys.stderr,

                    )

                    continue



                # Load metadata from the original file

                metadata = find_metadata_files(proc_fmt, infile)



            input_audio = fromfile(proc_fmt, infile, in_meta=metadata)

                # Measure input loudness

                input_audio = fromfile(

                    proc_fmt, resampled_file, fs=int(fs) * 1000, in_meta=metadata

                )

                input_loudness, _, input_loudness_format = get_loudness(input_audio)

            input_audio_cache[proc_fmt] = (input_loudness, input_loudness_format)



        input_loudness, input_loudness_format = input_audio_cache[proc_fmt]

                input_audio_cache[cache_key] = (input_loudness, input_loudness_format)



        # Find all output files for this format

        output_files = [
@@ -259,6 +279,20 @@ def main(args): 
        ]



        for outfile in output_files:

            # Get bandwidth from output filename to look up correct cached input

            try:

                _, _, _, bandwidth = get_metadata_from_outfile_name(

                    pattern, outfile, infile.stem

                )

            except:

                continue



            cache_key = (proc_fmt, bandwidth)

            if cache_key not in input_audio_cache:

                continue



            input_loudness, input_loudness_format = input_audio_cache[cache_key]



            tasks.append(

                (

                    outfile,