Parallel Audio Feature Extraction with MP.py

    for ebird in final_data:
        print("Starting to process a new species: ", ebird)
        ebird_data = train_csv[train_csv['species'] == ebird]

        short_file_name = ebird_data['ebird_code'].unique()[0]
        print("Short file name: ", short_file_name)

        pool = mp.Pool(c.NUMBER_OF_CPU_IN_POOL)  # use the number of parallel processes as per the configured

        funclist = []

        for index, row in ebird_data.iterrows():
            # process each audio file
            f = pool.apply_async(extract_feautres, [row['full_path']])
            funclist.append(f)

        result = []
        for f in funclist:
            result.append(f.get(timeout=600))  # timeout in 600 seconds = 10 mins

        # combine chunks with transformed data into a single training set
        extracted_features = pd.concat(result)

        # save extracted features to CSV
        output_path = "".join([c.TRANSFORMED_DATA_PATH, short_file_name, ".csv"])
        extracted_features.to_csv(output_path, index=False)

        # clean up
        pool.close()
        pool.join()

        print("Finished processing: ", ebird)