Merge branch 'dev_gzf_deepspeed' of http://gitlab.alibaba-inc.com/zhifu.gzf/FunASR into dev_gzf_deepspeed

funasr/metrics/compute_acc_age_gender.py

@@ -0,0 +1,106 @@
+# Copyright 3D-Speaker (https://github.com/alibaba-damo-academy/3D-Speaker). All Rights Reserved.
+# Licensed under the Apache License, Version 2.0 (http://www.apache.org/licenses/LICENSE-2.0)
+
+import os
+import argparse
+import numpy as np
+import pandas as pd
+
+# from speakerlab.utils.utils import get_utt2spk_dict
+
+parser = argparse.ArgumentParser(description="Output the score.")
+parser.add_argument("--predict", default="", type=str, help="Prediction results")
+parser.add_argument("--ground_truth", default="", type=str, help="Ground truth")
+parser.add_argument("--task", default="age", type=str, help="age, gender")
+parser.add_argument("--out_dir", default="", type=str, help="Out dir")
+
+
+def get_utt2spk_dict(utt2spk, suffix="", task="age"):
+    temp_dict = {}
+    with open(utt2spk, "r") as utt2spk_f:
+        lines = utt2spk_f.readlines()
+    for i in lines:
+        i = i.strip().split(maxsplit=1)
+        if suffix == "" or suffix is None:
+            key_i = i[0]
+            value_spk = i[1]
+        else:
+            key_i = i[0] + "_" + suffix
+            value_spk = i[1] + "_" + suffix
+        if task == "gender":
+            value_spk = value_spk.replace("性别:", "").strip()
+            value_spk = value_spk.replace("Gender:", "")
+            value_spk = value_spk.strip().lower()
+        elif task == "age":
+            pass
+        if key_i in temp_dict:
+            raise ValueError("The key must be unique.")
+        temp_dict[key_i] = value_spk
+    return temp_dict
+
+
+def main():
+    args = parser.parse_args()
+    predict = get_utt2spk_dict(args.predict)
+    ground = get_utt2spk_dict(args.ground_truth)
+    langs = list(set(ground.values()))
+    langs.sort()
+    # print('-------langs--------')
+    # print(langs) # ['fifties', 'seventies', 'sixties', 'thirties', 'fourties', 'eighties', 'twenties', 'teens']
+    confusion = np.zeros((len(langs), len(langs)))
+    sum = 0
+    cor = 0
+    for utt_id in predict:
+        sum += 1
+        confusion[langs.index(ground[utt_id])][langs.index(predict[utt_id])] += 1
+        if predict[utt_id] == ground[utt_id]:
+            cor += 1
+
+    os.makedirs(args.out_dir, exist_ok=True)
+    with open("%s/acc.res" % args.out_dir, "w") as f:
+        f.write("Acc:%.2f%%" % (cor / sum * 100))
+    print("Acc:%.2f%%" % (cor / sum * 100))
+    # print('-------confusion------')
+    # print(confusion)
+
+    writeExcel(confusion, langs, "%s/predict.xlsx" % args.out_dir)
+
+
+def writeExcel(data, langs, excel_path):
+    assert data.shape[0] == len(langs)
+    num = data.shape[0]
+    form = np.zeros((num + 1, num + 4))
+    for i in range(num):
+        form[num, i] = sum(data[:, i])
+    for i in range(num):
+        form[i, num] = sum(data[i])
+        form[i, num + 1] = data[i, i] / (float(form[i, num]) + 1e-10)
+        form[i, num + 2] = data[i, i] / (float(form[num, i]) + 1e-10)
+        form[i, num + 3] = (
+            2
+            * form[i, num + 1]
+            * form[i, num + 2]
+            / (float((form[i, num + 1] + form[i, num + 2])) + 1e-10)
+        )
+    form[:num, :num] = data
+    # print('---------form--------')
+    # print(form)
+    data_pd = pd.DataFrame(form)
+    data_pd.columns = langs + ["Total", "Recall", "Precision", "F1"]
+    data_pd.index = langs + ["Total"]
+    writer = pd.ExcelWriter(excel_path)
+    data_pd.to_excel(writer, float_format="%.3f")
+    writer.close()
+
+
+if __name__ == "__main__":
+    main()
+
+"""
+predict=
+ground_truth=
+task=
+out_dir=
+
+python funasr/metrics/compute_acc_age_gender.py --predict ${predict} --ground_truth ${ground_truth} --task ${task} --out_dir ${out_dir}
+"""