update

examples/sample_filter/test1.py

@@ -0,0 +1,69 @@
+#!/usr/bin/python3
+# -*- coding: utf-8 -*-
+import argparse
+from pathlib import Path
+import shutil
+
+from gradio_client import Client, handle_file
+from tqdm import tqdm
+
+
+def get_args():
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--data_dir",
+        default=r"E:\Users\tianx\HuggingDatasets\vm_sound_classification\data\temp\temp",
+        type=str
+    )
+    parser.add_argument(
+        "--keep_dir",
+        default=r"E:\Users\tianx\HuggingDatasets\vm_sound_classification\data\keep",
+        type=str
+    )
+    parser.add_argument(
+        "--trash_dir",
+        default=r"E:\Users\tianx\HuggingDatasets\vm_sound_classification\data\trash",
+        type=str
+    )
+    args = parser.parse_args()
+    return args
+
+
+def main():
+    args = get_args()
+
+    data_dir = Path(args.data_dir)
+    keep_dir = Path(args.keep_dir)
+    keep_dir.mkdir(parents=True, exist_ok=True)
+    trash_dir = Path(args.trash_dir)
+    trash_dir.mkdir(parents=True, exist_ok=True)
+
+    client = Client("http://127.0.0.1:7864/")
+
+    for filename in tqdm(data_dir.glob("*.wav")):
+        filename = filename.as_posix()
+
+        label, prob = client.predict(
+            audio=handle_file(filename),
+            model_name="vm_sound_classification8-ch32",
+            ground_true="Hello!!",
+            api_name="/click_button"
+        )
+        prob = float(prob)
+
+        if prob > 0.7:
+            shutil.move(
+                filename,
+                trash_dir.as_posix(),
+            )
+        else:
+            shutil.move(
+                filename,
+                keep_dir.as_posix(),
+            )
+
+    return
+
+
+if __name__ == '__main__':
+    main()

examples/vm_sound_classification/run.sh

@@ -15,7 +15,9 @@ E:/Users/tianx/HuggingDatasets/vm_sound_classification/data/wav_finished/id-ID/w
 sh run.sh --stage 0 --stop_stage 5 --system_version centos --file_folder_name file_dir --final_model_name vm_sound_classification8-ch32 \
 --filename_patterns "/data/tianxing/PycharmProjects/datasets/voicemail/*/wav_finished/*/*.wav" --label_plan 8
 
-"
+sh run.sh --stage 0 --stop_stage 5 --system_version centos --file_folder_name file_dir --final_model_name vm_sound_classification2-ch32-voicemail \
+--filename_patterns "/data/tianxing/PycharmProjects/datasets/voicemail/th-TH/wav_finished/*/*.wav" --label_plan 2-voicemail
+
 
 END
 

examples/vm_sound_classification/step_1_prepare_data.py

@@ -39,7 +39,20 @@ def get_dataset(args):
     file_dir = Path(args.file_dir)
     file_dir.mkdir(exist_ok=True)
 
-    if args.label_plan == "2":
+    if args.label_plan == "2-voicemail":
+        label_map = {
+            "bell": "voicemail",
+            "white_noise": "non_voicemail",
+            "low_white_noise": "non_voicemail",
+            "high_white_noise": "non_voicemail",
+            # "music": "non_voicemail",
+            "mute": "non_voicemail",
+            "noise": "non_voicemail",
+            "noise_mute": "non_voicemail",
+            "voice": "non_voicemail",
+            "voicemail": "voicemail",
+        }
+    elif args.label_plan == "2":
         label_map = {
             "bell": "non_voice",
             "white_noise": "non_voice",