Add a tiny visualization tool

utils.py

@@ -60,7 +60,8 @@ def get_speech_ts(wav: torch.Tensor,
                   batch_size: int = 200,
                   num_samples_per_window: int = 4000,
                   min_speech_samples: int = 10000, #samples
-                  run_function=validate):
+                  run_function=validate,
+                  visualize_probs=False):
 
     num_samples = num_samples_per_window
     assert num_samples % num_steps == 0
@@ -89,14 +90,20 @@ def get_speech_ts(wav: torch.Tensor,
     triggered = False
     speeches = []
     current_speech = {}
+    if visualize_probs:
+      import pandas as pd
+      smoothed_probs = []
 
     speech_probs = outs[:, 1]  # this is very misleading
     for i, predict in enumerate(speech_probs):  # add name
         buffer.append(predict)
-        if ((sum(buffer) / len(buffer))>= trig_sum) and not triggered:
+        smoothed_prob = (sum(buffer) / len(buffer))
+        if visualize_probs:
+          smoothed_probs.append(float(smoothed_prob))
+        if (smoothed_prob >= trig_sum) and not triggered:
             triggered = True
             current_speech['start'] = step * max(0, i-num_steps)
-        if ((sum(buffer) / len(buffer)) < neg_trig_sum) and triggered:
+        if (smoothed_prob < neg_trig_sum) and triggered:
             current_speech['end'] = step * i
             if (current_speech['end'] - current_speech['start']) > min_speech_samples:
                 speeches.append(current_speech)
@@ -105,6 +112,9 @@ def get_speech_ts(wav: torch.Tensor,
     if current_speech:
         current_speech['end'] = len(wav)
         speeches.append(current_speech)
+    
+    if visualize_probs:
+      pd.DataFrame({'probs':smoothed_probs}).plot(figsize=(16,8))
     return speeches