f5_tuned_results/overall.py

import pandas as pd

def main():
    nisqa_df = pd.read_csv("NISQA_results.csv")

    noisiness = nisqa_df["noi_pred"].mean()
    coloration = nisqa_df["col_pred"].mean()
    discontinuity = nisqa_df["dis_pred"].mean()
    loudness = nisqa_df["loud_pred"].mean()
    mos_v1 = nisqa_df["mos_pred"].mean()

    mos_df = pd.read_csv("NISQA_results_mos.csv")
    mos_v2 = mos_df["mos_pred"].mean()

    cossim_df = pd.read_csv("cossim.csv")
    cossim = cossim_df["cossim"].mean()

    metrics_df = pd.read_csv("metadata.csv", sep="|")
    metrics_df.columns = ["uttr", "rtf", "ttfa"]
    avg_row = metrics_df[metrics_df["uttr"] == "average"]

    rtf = avg_row["rtf"].iloc[0]
    ttfa = avg_row["ttfa"].iloc[0] * 1000

    l = [cossim, mos_v2, mos_v1, noisiness, discontinuity, coloration, loudness, ttfa, "-", rtf]
    _l = list(map(lambda x: str(x), l))

    print(",".join(_l))

if __name__ == "__main__":
    main()