Update app.py

app.py

@@ -21,6 +21,8 @@ from torchmetrics.audio.pesq import PerceptualEvaluationSpeechQuality as PESQ
 
 
 from PLCMOS.plc_mos import PLCMOSEstimator
+from speechmos import dnsmos
+from speechmos import plcmos
 
 
 @st.cache
@@ -226,29 +228,33 @@ if st.button('Сгенерировать потери'):
     
 
 
-    pesq_orig = pesq(fs = 16000, ref = data_clean, deg = data_clean, mode='wb')
-    pesq_lossy = pesq(fs = 16000, ref = data_clean, deg = data_lossy, mode='wb')
-    pesq_enhanced = pesq(fs = 16000, ref = data_clean, deg = data_enhanced, mode='wb')
+    pesq_orig = pesq(fs = 8000, ref = data_clean, deg = data_clean, mode='nb')
+    pesq_lossy = pesq(fs = 8000, ref = data_clean, deg = data_lossy, mode='nb')
+    pesq_enhanced = pesq(fs = 8000, ref = data_clean, deg = data_enhanced, mode='nb')
 
     psq_mas=[pesq_orig, pesq_lossy, pesq_enhanced]
 
     
 
+    data_clean, fs = sf.read('target.wav')
+    data_lossy, fs = sf.read('lossy.wav')
+    data_enhanced, fs = sf.read('enhanced.wav')
 
-
-
-
+    if fs!= 16000:
+        data_lossy = librosa.resample(data_lossy, orig_sr=48000, target_sr=16000)
+        data_clean = librosa.resample(data_clean, orig_sr=48000, target_sr=16000)
+        data_enhanced = librosa.resample(data_enhanced, orig_sr=48000, target_sr=16000)
 
     PLC_example=PLCMOSEstimator()
     PLC_org = PLC_example.run(audio_degraded=data_clean, audio_clean=data_clean)[0]
     PLC_lossy = PLC_example.run(audio_degraded=data_lossy, audio_clean=data_clean)[0]
     PLC_enhanced = PLC_example.run(audio_degraded=data_enhanced, audio_clean=data_clean)[0]
 
-    PLC_mass = [PLC_org, PLC_lossy, PLC_enhanced]
+    PLC_massv1 = [PLC_org, PLC_lossy, PLC_enhanced]
 
 
     
-    df = pd.DataFrame(columns=['Audio', 'PESQ', 'STOI', 'PLCMOS', 'LSD'])
+    df = pd.DataFrame(columns=['Audio', 'PESQ', 'STOI', 'PLCMOSv1', 'DNSMOS', 'PLCMOSv2'])
 
     df['Audio'] = ['Clean', 'Lossy', 'Enhanced']
 
@@ -256,10 +262,18 @@ if st.button('Сгенерировать потери'):
 
     df['STOI'] = stoi_mass
 
-    df['LSD'] = lsd_mass
+    #df['LSD'] = lsd_mass
 
-    df['PLCMOS'] = PLC_mass
+    df['PLCMOSv1'] = PLC_massv1
 
+
+    PLC_massvV2 = [plcmos.run("target.wav", sr=16000)['plcmos'], plcmos.run("lossy.wav", sr=16000)['plcmos'], plcmos.run("enhanced.wav", sr=16000)['plcmos']]
+
+    df['PLCMOSv2'] = PLC_massv2
+
+    df.columns = pd.MultiIndex.from_tuples(zip(['', 'Intrusive metrics', '', '', 'Non intrusive metrics', ''], df.columns))
+
+    
     st.table(df)