Update host.py

host.py

@@ -7,6 +7,7 @@ import librosa
 from scipy.spatial.distance import cosine
 import numpy as np
 import os
+import time
 # brew install ffmpeg
 # pip install flask transformers librosa torch torchaudio
 
@@ -43,6 +44,15 @@ def preprocess_audio(audio_data):
     waveform = waveform.squeeze().numpy()
     return waveform
 
+
+# Preload expected audio
+mp3_audio = preprocess_audio(mp3_file_path)
+mp3_inputs = feature_extractor(mp3_audio, return_tensors="pt")
+mp3_embeddings = model(**mp3_inputs).embeddings
+mp3_embeddings_normalized = torch.nn.functional.normalize(mp3_embeddings, dim=-1).cpu()
+
+mp3_audio2 = open(mp3_file_path2, 'rb').read()
+
 @app.route('/')
 def index():
     return render_template('index.html')
@@ -58,6 +68,7 @@ def chal2():
 @app.route('/compare_audio', methods=['POST'])
 def compare_audio():
     try:
+        start_time = time.time()
         # Get the recorded audio file from the frontend
         recorded_audio = request.files['audio_data']
 
@@ -68,22 +79,20 @@ def compare_audio():
         embeddings_normalized = torch.nn.functional.normalize(embeddings, dim=-1).cpu()
 
         # Load and preprocess MP3 file for comparison
-        mp3_audio = preprocess_audio(mp3_file_path)
-        mp3_inputs = feature_extractor(mp3_audio, return_tensors="pt")
-        mp3_embeddings = model(**mp3_inputs).embeddings
-        mp3_embeddings_normalized = torch.nn.functional.normalize(mp3_embeddings, dim=-1).cpu()
+        global mp3_embeddings_normalized
 
         # Calculate cosine similarity
         cosine_sim = torch.nn.CosineSimilarity(dim=-1)
         similarity = cosine_sim(embeddings_normalized, mp3_embeddings_normalized).item()
 
         similarity = round(similarity, 3)
+        end = time.time()-start_time
 
         threshold = 0.89  # Adjust the threshold as needed
         if similarity < threshold:
-            result = "Authorization Failed! " + str(similarity) + " < 0.890<br>Do your best Terminator impression"
+            result = "Authorization Failed! " + str(similarity) + " < 0.890"+" in "+str(round(end,3))+"s"+"<br>Do your best Terminator impression"
         else:
-            result = "Good job! Match: " + str(similarity) + "<br>" + flag1 + "<br><a href='/chal2'>Click here to open the next challenge</a>"
+            result = "Good job! Match: " + str(similarity) + "<br>" + flag1 + "<br><a href='/chal2'>Click here to open the next challenge</a>"+"<br>processed in "+str(round(end,3))+"s"
 
         return jsonify({'result': result})
     except Exception as e:
@@ -118,15 +127,18 @@ def preprocess_audio2(audio_bytes):
 
     return waveform
 
+mfcc2 = extract_mfcc(mp3_audio2)
+
 @app.route('/compare_audio2', methods=['POST'])
 def compare_audio2():
     try:
         recorded_audio = request.files['audio_data'].read()
-        mp3_audio = open(mp3_file_path2, 'rb').read()
+        #mp3_audio = open(mp3_file_path2, 'rb').read()
 
         # Compare similarity between audio
         mfcc1 = extract_mfcc(recorded_audio)
-        mfcc2 = extract_mfcc(mp3_audio)
+        #mfcc2 = extract_mfcc(mp3_audio)
+        global mfcc2
         similarity = 1 - cosine(np.mean(mfcc1, axis=1), np.mean(mfcc2, axis=1))
         similarity = round(similarity, 3)
         if similarity < 0.940:
@@ -140,4 +152,4 @@ def compare_audio2():
         return jsonify({'error': 'An error occurred during audio comparison. Im fragile please dont abuse.'})
 
 if __name__ == '__main__':
-    app.run(host="0.0.0.0", port=8080, debug=True)
+    app.run(host="0.0.0.0", port=8080, debug=True)