Merge branch 'master' of github.com:bagustris/nkululeko

data/jnv/README.md

@@ -0,0 +1,12 @@
+# JNV Corpus  
+
+Pre-processing the JNV Corpus for Nkululeko (CSV format).
+
+```bash
+wget https://ss-takashi.sakura.ne.jp/corpus/jnv/jnv_corpus_ver2.zip
+unzip jnv_corpus_ver2.zip
+cd ../..
+# the following will resample and replace JNV forpus to 16k 
+python3 -m nkululeko.resample --config data/jnv/exp.ini
+python3 -m nkululeko.nkululeko --config data/jnv/exp.ini
+```

data/jnv/process_database.py

@@ -0,0 +1,42 @@
+# process_database.py: pre-processing script for JNV database
+
+import argparse
+import json
+import os
+import sys
+
+import numpy as np
+import pandas as pd
+from tqdm import tqdm
+
+
+def read_audio_files(data_dir):
+    data = []
+    for root, dirs, files in os.walk(data_dir):
+        for file in files:
+            if file.endswith(".wav"):
+                emotion = file.split("_")[1]
+                data.append({"file": os.path.join(root, file), "emotion": emotion})
+
+    df = pd.DataFrame(data)
+    return df
+
+def main(args):
+    data_dir = args.data_dir
+    output_file = args.output_file
+
+    df = read_audio_files(data_dir)
+    train_df, dev_df, test_df = np.split(df.sample(frac=1, random_state=42), [int(.8*len(df)), int(.9*len(df))])
+
+    train_df.to_csv("jnv_train.csv", index=False)
+    dev_df.to_csv("jnv_dev.csv", index=False)  
+    test_df.to_csv("jnv_test.csv", index=False)
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--data_dir", type=str, default="jnv_corpus_ver2/JNV/", help="Directory containing audio files")
+    parser.add_argument("--output_file", type=str, default="jnv_database.csv", help="Output CSV file")
+    args = parser.parse_args()
+
+    main(args)