update

examples/vm_sound_classification/run.sh

@@ -2,13 +2,14 @@
 
 : <<'END'
 
-sh run.sh --stage 3 --stop_stage 3 --system_version windows --file_folder_name file_dir --final_model_name vm_sound_classification3 \
---filename_patterns "E:/programmer/asr_datasets/voicemail/wav_finished/en-US/wav_finished/*/*.wav \
-E:/programmer/asr_datasets/voicemail/wav_finished/id-ID/wav_finished/*/*.wav" \
+sh run.sh --stage 0 --stop_stage 1 --system_version windows --file_folder_name file_dir --final_model_name vm_sound_classification4-ch16 \
+--filename_patterns "E:/Users/tianx/HuggingDatasets/vm_sound_classification/data/wav_finished/wav_finished/en-US/wav_finished/*/*.wav \
+E:/Users/tianx/HuggingDatasets/vm_sound_classification/data/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_classification4-ch16 \
+--filename_patterns "E:/Users/tianx/HuggingDatasets/vm_sound_classification/data/wav_finished/wav_finished/en-US/wav_finished/*/*.wav \
+E:/Users/tianx/HuggingDatasets/vm_sound_classification/data/wav_finished/id-ID/wav_finished/*/*.wav" \
 
-sh run.sh --stage 0 --stop_stage 1 --system_version windows --file_folder_name file_dir --final_model_name vm_sound_classification3 \
---filename_patterns "E:/programmer/asr_datasets/voicemail/wav_finished/en-US/wav_finished/*/*.wav \
-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

examples/vm_sound_classification/step_3_train_model.py

@@ -83,6 +83,13 @@ class CollateFunction(object):
             array = sample["waveform"]
             label = sample["label"]
 
+            l = len(array)
+            if l < 16000:
+                delta = int(16000 - l)
+                array = np.concatenate([array, np.zeros(shape=(delta,), dtype=np.float32)], axis=-1)
+            if l > 16000:
+                array = array[:16000]
+
             array_list.append(array)
             label_list.append(label)