update

examples/vm_sound_classification/conv2d_classifier.yaml

@@ -13,21 +13,21 @@ mel_spectrogram_param:
 conv2d_block_param_list:
   - batch_norm: true
     in_channels: 1
-    out_channels: 16
+    out_channels: 8
     kernel_size: 3
     stride: 1
     dilation: 3
     activation: relu
     dropout: 0.1
-  - in_channels: 16
-    out_channels: 16
+  - in_channels: 8
+    out_channels: 8
     kernel_size: 5
     stride: 2
     dilation: 3
     activation: relu
     dropout: 0.1
-  - in_channels: 16
-    out_channels: 16
+  - in_channels: 8
+    out_channels: 8
     kernel_size: 3
     stride: 1
     dilation: 2

examples/vm_sound_classification/run.sh

@@ -13,7 +13,7 @@ E:/programmer/asr_datasets/voicemail/wav_finished/id-ID/wav_finished/*/*.wav" \
 sh run.sh --stage 2 --stop_stage 2 --system_version windows --file_folder_name file_dir --final_model_name vm_sound_classification3
 sh run.sh --stage 3 --stop_stage 3 --system_version windows --file_folder_name file_dir --final_model_name vm_sound_classification3
 
-sh run.sh --stage 0 --stop_stage 1 --system_version centos --file_folder_name file_dir --final_model_name vm_sound_classification8-ch16 \
+sh run.sh --stage 3 --stop_stage 5 --system_version centos --file_folder_name file_dir --final_model_name vm_sound_classification8-ch8 \
 --filename_patterns "/data/tianxing/PycharmProjects/datasets/voicemail/*/wav_finished/*/*.wav"
 
 

main.py

@@ -105,24 +105,23 @@ def main():
     examples_dir = Path(args.examples_dir)
     trained_model_dir = Path(args.trained_model_dir)
 
+    # models
+    model_choices = list()
+    for filename in trained_model_dir.glob("*.zip"):
+        model_name = filename.stem
+        model_choices.append(model_name)
+
     # examples
     examples = list()
     for filename in examples_dir.glob("*/*/*.wav"):
-        language = filename.parts[-3]
         label = filename.parts[-2]
 
         examples.append([
             filename.as_posix(),
-            language,
+            model_choices[0],
             label
         ])
 
-    # models
-    model_choices = list()
-    for filename in trained_model_dir.glob("*.zip"):
-        model_name = filename.stem
-        model_choices.append(model_name)
-
     # ui
     brief_description = """
 国际语音智能外呼系统, 电话声音分类. 
@@ -137,7 +136,7 @@ def main():
                 c_audio = gr.Audio(label="audio")
                 with gr.Row():
                     with gr.Column(scale=3):
-                        c_model_name = gr.Dropdown(choices=model_choices, value=model_choices[0], label="language")
+                        c_model_name = gr.Dropdown(choices=model_choices, value=model_choices[0], label="model_name")
                     with gr.Column(scale=3):
                         c_ground_true = gr.Textbox(label="ground_true")