add librosa to requirements and add audio labelmap

pipeline.py

@@ -205,6 +205,11 @@ def load_audio_model():
     model = model.load_state_dict(ckpt, model_dict)
     return model
 
+audio_label_map = {
+    0: "real"
+    1: "fake"
+}
+
 def deepfakes_audio_predict(input_audio):
     #Perform inference on audio.
     x, sr = librosa.load(input_audio)
@@ -220,4 +225,5 @@ def deepfakes_audio_predict(input_audio):
     #Get the argmax.
     grads_np = grads.detach().numpy()
     result = np.argmax(grads_np)
-    return result
+
+    return audio_label_map[result]

requirements.txt

@@ -5,4 +5,5 @@ numpy
 opencv-python
 opencv-python-headless
 mtcnn
-moviepy
+moviepy
+librosa