updated files

app.py

@@ -1,54 +1,13 @@
-
-import numpy as np
-import tensorflow as tf
-from scipy.io.wavfile import write
-import keras.backend as K
-import librosa.display
-import cv2
-import librosa
-import matplotlib.pyplot as plt
-import librosa.display
-import numpy as np
-from keras.applications import VGG16
 import os
-import scipy
 import gradio as gr
 import shutil
+from main_code import main_loop
+
 
-# Load the tune recognition model
-model = tf.keras.models.load_model('embdmodel_1.hdf5')
-embedding_model=model.layers[2]
 
 DURATION = 10
 WAVE_OUTPUT_FILE = "my_audio.wav"
 
-
-# Define function to preprocess input audio
-#convert song to mel spectogram as siamese network doesn't work on sound directly
-def create_spectrogram(clip,sample_rate,save_path):
-    plt.interactive(False)
-    fig=plt.figure(figsize=[0.72,0.72])
-    S=librosa.feature.melspectrogram(y=clip,sr=sample_rate)
-    librosa.display.specshow(librosa.power_to_db(S,ref=np.max))
-    fig.savefig(save_path,dpi=400,bbox_inches='tight',pad_inches=0)
-    plt.close()
-    fig.clf()
-    plt.close(fig)
-    plt.close('all')
-    del save_path,clip,sample_rate,fig,S
-    
-def load_img(path):
-    img=cv2.imread(path)
-    img=cv2.cvtColor(img,cv2.COLOR_BGR2RGB)
-    img=cv2.resize(img,(150,150))
-    return img
-
-
-import pickle
-with open('dict.pickle', 'rb') as handle:
-    songspecdict = pickle.load(handle)
-
-
 def list_file_sizes():
     path = "."
  
@@ -76,75 +35,9 @@ def main(audio):
 
     list_file_sizes()
 
-    # Load the song to match
-    song, sr = librosa.load("my_audio.wav")
-    to_match = np.copy(song[0:220500])
-    print("Loaded data into librosa...")
-
-    # Create spectrogram image of the song to match
-    create_spectrogram(to_match, sr, 'test.png')
-    print("Created spectogram...")
-
-    # Load the spectrogram image of the song to match
-    to_match_img = load_img('test.png')
-    to_match_img = np.expand_dims(to_match_img, axis=0)
-    print("Loaded spectrum image...")
-
-    # Get the embedding of the song to match
-    to_match_emb = embedding_model.predict(to_match_img)
-    print("Get song embedding...")
-
-    # Calculate the distances between the song to match and the songs in the database
-    songsdistdict = {}
-    for key, values in songspecdict.items():
-        dist_array = []
-        for embd in values:
-            dist_array.append(np.linalg.norm(to_match_emb - embd))
-            
-        songsdistdict[key] = min(dist_array)
-    song_titles=list(songsdistdict.keys())
-    distances=list(songsdistdict.values())
-
-    # Get the title and artist of the recognized song
-    recognized_song_artist, recognized_song_title = song_titles[distances.index(min(distances))].split('-')
-    recognized_song_title = os.path.splitext(recognized_song_title)[0]
-    print(f'Artist: {recognized_song_artist}')
-    print(f'Title: {recognized_song_title}')
-
-    from musixmatch import Musixmatch
-
-    # Initialize Musixmatch API
-    musixmatch = Musixmatch(apikey='2b0d0615efa782e95598a0e99bda4a60')
-
-    # Search for the recognized song
-    track_search_results = musixmatch.track_search(q_track=recognized_song_title, q_artist=recognized_song_artist, page_size=1, page=1, s_track_rating='desc')
-
-    if track_search_results['message']['header']['status_code'] == 200:
-        # Get the track ID for the top result
-        track_id = track_search_results['message']['body']['track_list'][0]['track']['track_id']
-
-        # Get the lyrics for the recognized song
-        lyrics_result = musixmatch.track_lyrics_get(track_id=track_id)
-
-        if lyrics_result['message']['header']['status_code'] == 200:
-            # Get the lyrics
-            lyrics = lyrics_result['message']['body']['lyrics']['lyrics_body']
-            # Remove the annotation tags from the lyrics
-            lyrics = lyrics.replace('******* This Lyrics is NOT for Commercial use *******', '').strip()
-            print("Lyrics:\n", lyrics)
-    else:
-        print("Couldn't find lyrics for the recognized song.")
-
-
-
-    # Play the recognized song
-    recognized_song_file = f'https://huggingface.co/spaces/prerna9811/Chord/tree/main/seismese_net_songs/{song_titles[distances.index(min(distances))]}'
-    recognized_song_audio, recognized_song_sr = librosa.load(recognized_song_file)
-
-    audio_file = open(recognized_song_file, 'rb') # enter the filename with filepath
-    audio_bytes = audio_file.read() # reading the file
+    song = main_loop()
 
-    return audio_bytes
+    return audio
 
 
 import asyncio
@@ -155,7 +48,7 @@ demo = gr.Blocks()
 mf_transcribe = gr.Interface(
             fn=main,
             inputs=gr.inputs.Audio(source="microphone", type="filepath"),
-            outputs="audio",
+            outputs="text",
             layout="horizontal",
             theme="huggingface",
         )