Update app.py

app.py

@@ -1,56 +1,74 @@
-import tensorflow as tf
-import gradio as gr
 import librosa
 import numpy as np
-import json
+import tensorflow as tf
+import gradio as gr
 
-# File Paths
-model_path = "audio_clf_model"
-encoding_path = "label_encodings.json"
-examples_path = "examples"
 
-# loading the files
+# File Paths
+model_path = "sound_emotion_rec_model"
+categories = ['angry', 'disgust', 'fear', 'happy', 'neutral', 'ps', 'sad']
 model = tf.keras.models.load_model(model_path)
-classes = json.load(open(encoding_path, "r"))
-labels = [classes[str(i)] for i in range(len(classes))]
-labels = ["negative", "positive"]#classes[target].values()
 
-def pre_processor(audio_path):
 
-  # load the audio file
-  x, sample_rate = librosa.load(audio_path)
-  
-  # feature extracting (mfccs is an aduio feature)
-  mfccs = np.mean(librosa.feature.mfcc(y=x, sr=sample_rate, n_mfcc=40).T, axis=0)
-  feature = mfccs
+# loading the files
+def extract_mfcc(audio_path, duration=3, offset=0.5, n_mfcc=40):
+    # loading the data
+    y, sr = librosa.load(audio_path, duration=duration, offset=offset)
+    
+    # extracting the voice feature
+    mfcc = np.mean(librosa.feature.mfcc(y=y, sr=sr, n_mfcc=n_mfcc).T, axis=0)
+    
+    return mfcc
+
+def prepare_data(audio_path):
   
-  return feature
+  # extracting the features
+  features = extract_mfcc(audio_path)
+
+  # adjusting the shape
+  features = [x for x in features]
+  features = np.array(features)
+  features = np.expand_dims(features, -1)
 
+  return features
 
 def clsf(audio_path):
   
   # extracting the features
-  features = pre_processor(audio_path)
-  print(len(features))
+  features = prepare_data(audio_path)
 
   # batching the data
   sample = np.expand_dims(features, axis=0)
 
   # predicting
-  preds = model.predict(sample).flatten()
-  
+  preds = model.predict(sample)[0]
+
   # results
-  confidences = {labels[i]: np.round(float(preds[i]), 3) for i in range(len(labels))}
+  confidences = {categories[i]:np.round(float(preds[i]), 3) for i in range(len(categories))}
 
   return confidences
 
+def pre_processor(audio_path):
+
+  # load the audio file
+  x, sample_rate = librosa.load(audio_path)
+  
+  # feature extracting (mfccs is an aduio feature)
+  mfccs = np.mean(librosa.feature.mfcc(y=x, sr=sample_rate, n_mfcc=40).T, axis=0)
+  feature = mfccs
+  
+  return feature
+
+
+
 # GUI Component
 gui_params = {
     "fn":clsf, 
     "inputs":gr.Audio(source="upload", type="filepath"),
     "outputs" : "label",
     #live=True,
-    "examples" : examples_path
+    "examples" : "examples"
+    
 } 
 demo = gr.Interface(**gui_params)