Values

app.py

@@ -0,0 +1,21 @@
+import streamlit as st
+import librosa
+from src.preprosecing import preprosecing
+st.write("Clasificación de canciones")
+#st.set_page_config(page_title="Upload Music", page_icon=":musical_note:", layout="wide")
+
+def main():
+    uploaded_file = st.file_uploader("Choose a music file", type=["mp3"])
+
+    if uploaded_file is not None:
+        uploaded_file, features = preprosecing(uploaded_file)
+        st.audio(uploaded_file, format='audio/wav')
+        st.success("30 secs audio snippet")
+        st.success("File uploaded successfully")
+        st.write("This is the features from the audio")
+        st.write(features)
+    else:
+        st.warning("Please upload a file of type: mp3, wav")
+
+if __name__ == "__main__":
+    main()

requierements.in

@@ -0,0 +1,10 @@
+sklearn
+pandas
+librosa
+matplotlib
+lightgbm
+catboost
+xgboost
+streamlit
+seaborn
+pydub

requierements.txt

@@ -0,0 +1,213 @@
+#
+# This file is autogenerated by pip-compile with Python 3.10
+# by the following command:
+#
+#    pip-compile requierements.in
+#
+altair==4.2.0
+    # via streamlit
+appdirs==1.4.4
+    # via pooch
+attrs==22.2.0
+    # via jsonschema
+audioread==3.0.0
+    # via librosa
+blinker==1.5
+    # via streamlit
+cachetools==5.2.1
+    # via streamlit
+catboost==1.1.1
+    # via -r requierements.in
+certifi==2022.12.7
+    # via requests
+cffi==1.15.1
+    # via soundfile
+charset-normalizer==2.1.1
+    # via requests
+click==8.1.3
+    # via streamlit
+commonmark==0.9.1
+    # via rich
+contourpy==1.0.6
+    # via matplotlib
+cycler==0.11.0
+    # via matplotlib
+decorator==5.1.1
+    # via
+    #   librosa
+    #   validators
+entrypoints==0.4
+    # via altair
+fonttools==4.38.0
+    # via matplotlib
+gitdb==4.0.10
+    # via gitpython
+gitpython==3.1.30
+    # via streamlit
+graphviz==0.20.1
+    # via catboost
+idna==3.4
+    # via requests
+importlib-metadata==6.0.0
+    # via streamlit
+jinja2==3.1.2
+    # via
+    #   altair
+    #   pydeck
+joblib==1.2.0
+    # via
+    #   librosa
+    #   scikit-learn
+jsonschema==4.17.3
+    # via altair
+kiwisolver==1.4.4
+    # via matplotlib
+librosa==0.9.2
+    # via -r requierements.in
+lightgbm==3.3.4
+    # via -r requierements.in
+llvmlite==0.39.1
+    # via numba
+markupsafe==2.1.1
+    # via jinja2
+matplotlib==3.6.2
+    # via
+    #   -r requierements.in
+    #   catboost
+    #   seaborn
+numba==0.56.4
+    # via
+    #   librosa
+    #   resampy
+numpy==1.23.5
+    # via
+    #   altair
+    #   catboost
+    #   contourpy
+    #   librosa
+    #   lightgbm
+    #   matplotlib
+    #   numba
+    #   pandas
+    #   pyarrow
+    #   pydeck
+    #   resampy
+    #   scikit-learn
+    #   scipy
+    #   seaborn
+    #   streamlit
+    #   xgboost
+packaging==23.0
+    # via
+    #   librosa
+    #   matplotlib
+    #   pooch
+    #   streamlit
+pandas==1.5.2
+    # via
+    #   -r requierements.in
+    #   altair
+    #   catboost
+    #   seaborn
+    #   streamlit
+pillow==9.4.0
+    # via
+    #   matplotlib
+    #   streamlit
+plotly==5.11.0
+    # via catboost
+pooch==1.6.0
+    # via librosa
+protobuf==3.20.3
+    # via streamlit
+pyarrow==10.0.1
+    # via streamlit
+pycparser==2.21
+    # via cffi
+pydeck==0.8.0
+    # via streamlit
+pydub==0.25.1
+    # via -r requierements.in
+pygments==2.14.0
+    # via rich
+pympler==1.0.1
+    # via streamlit
+pyparsing==3.0.9
+    # via matplotlib
+pyrsistent==0.19.3
+    # via jsonschema
+python-dateutil==2.8.2
+    # via
+    #   matplotlib
+    #   pandas
+    #   streamlit
+pytz==2022.7
+    # via pandas
+pytz-deprecation-shim==0.1.0.post0
+    # via tzlocal
+requests==2.28.1
+    # via
+    #   pooch
+    #   streamlit
+resampy==0.4.2
+    # via librosa
+rich==13.0.1
+    # via streamlit
+scikit-learn==1.2.0
+    # via
+    #   librosa
+    #   lightgbm
+scipy==1.10.0
+    # via
+    #   catboost
+    #   librosa
+    #   lightgbm
+    #   scikit-learn
+    #   xgboost
+seaborn==0.12.2
+    # via -r requierements.in
+semver==2.13.0
+    # via streamlit
+six==1.16.0
+    # via
+    #   catboost
+    #   python-dateutil
+sklearn==0.0.post1
+    # via -r requierements.in
+smmap==5.0.0
+    # via gitdb
+soundfile==0.11.0
+    # via librosa
+streamlit==1.16.0
+    # via -r requierements.in
+tenacity==8.1.0
+    # via plotly
+threadpoolctl==3.1.0
+    # via scikit-learn
+toml==0.10.2
+    # via streamlit
+toolz==0.12.0
+    # via altair
+tornado==6.2
+    # via streamlit
+typing-extensions==4.4.0
+    # via streamlit
+tzdata==2022.7
+    # via pytz-deprecation-shim
+tzlocal==4.2
+    # via streamlit
+urllib3==1.26.14
+    # via requests
+validators==0.20.0
+    # via streamlit
+watchdog==2.2.1
+    # via streamlit
+wheel==0.38.4
+    # via lightgbm
+xgboost==1.7.3
+    # via -r requierements.in
+zipp==3.11.0
+    # via importlib-metadata
+
+# The following packages are considered to be unsafe in a requirements file:
+# setuptools

src/models/model.joblib

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b625bc7ac843c7f18ea2aacbce537b35d07051b75c1c8d84f2ef11c474a84d0b
+size 56772779

src/models/model.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5b7546f95b16cf021c16f5ff5d3b3e7b91cc294a126e5477a827c4be1dba4b8
+size 56775478

src/preprosecing.py

@@ -0,0 +1,235 @@
+import os
+import numpy as np
+import librosa
+import soundfile as sf
+
+
+import statistics as st
+from joblib import load
+from pydub import AudioSegment
+
+CLASSES = ['blues', 'classical', 'country', 'disco', 'hiphop', 'jazz', 'metal', 'pop', 'reggae', 'rock']
+
+class Features:
+    def __init__(self, y, sr, hop_length=5000):
+        """
+        Initialize the class with audio signal, sr and hop_length
+        :param y: audio signal
+        :param sr: sample rate of audio signal
+        :param hop_length: hop_length  parameter used while calculating the chroma_stft feature
+        """
+        self.y = np.split(y, 10)
+        self.sr = sr
+        self.hop_length = hop_length
+
+    def get_mean_var(self, y):
+        """
+        Helper function to get mean and variance of feature
+        :param y: audio feature
+        :return: mean, variance
+        """
+        mean = y.mean()
+        var = y.var()
+        return mean, var
+
+    def zero_crossing_rate(self, y):
+        """
+        Returns the zero-crossing rate of the audio signal
+        :return: mean and variance of zero-crossing rate
+        """
+        values = librosa.feature.zero_crossing_rate(y)
+        return self.get_mean_var(values)
+
+    def harmonic_and_per(self, y):
+        """
+        separates the harmonic and percussive components of the audio signal
+        :return: harmonic and percussive components' mean and variance
+        """
+        y_harm, y_perc  = librosa.effects.hpss(y)
+        harm = self.get_mean_var(y_harm)
+        perc = self.get_mean_var(y_perc)
+        return harm, perc
+    
+  
+    def tempo(self, y):
+            """
+            Extracts the tempo (beats per minute) of an audio signal.
+            
+            Parameters:
+                y (ndarray): The audio signal represented as an numpy array.
+            
+            Returns:
+                float: The tempo of the audio signal in beats per minute.
+            """
+            tempo = librosa.beat.tempo(y, sr=self.sr)
+            return tempo
+
+    def centroid(self, y):
+            """
+            Extracts the spectral centroid of an audio signal.
+            
+            Parameters:
+                y (ndarray): The audio signal represented as an numpy array.
+            
+            Returns:
+                tuple: A tuple containing the mean and variance of the spectral centroid.
+            """
+            centroid = librosa.feature.spectral_centroid(y, sr=self.sr)
+            return self.get_mean_var(centroid)
+    
+
+    
+    def mfccs(self, y):
+        """
+        Extracts the Mel-Frequency Cepstral Coefficients (MFCCs) of an audio signal.
+        
+        Parameters:
+            y (ndarray): The audio signal represented as an numpy array.
+        
+        Returns:
+            ndarray: An array containing the mean and variance of the MFCCs.
+        """
+        mfccs = librosa.feature.mfcc(y, sr=self.sr)
+        mean = mfccs.mean(axis=1)
+        var = mfccs.var(axis=1)
+        values = [[mean[i], var[i]] for i in range(mean.shape[0])]
+        return np.array(values).reshape(-1)
+    
+    def chroma_stft(self, y):
+        """
+        Extracts the chroma feature of an audio signal.
+        
+        Parameters:
+            y (ndarray): The audio signal represented as an numpy array.
+        
+        Returns:
+            tuple: A tuple containing the mean and variance of the chroma feature.
+        """
+        chroma = librosa.feature.chroma_stft(y, sr=self.sr, hop_length=self.hop_length)
+        return self.get_mean_var(chroma)
+    
+    def spectral_bandwidth(self, y):
+        """
+        Extracts the spectral bandwidth of an audio signal.
+        
+        Parameters:
+            y (ndarray): The audio signal represented as an numpy array.
+        
+        Returns:
+            tuple: A tuple containing the mean and variance of the spectral bandwidth.
+        """
+        spd = librosa.feature.spectral_bandwidth(y,sr=self.sr )
+        return self.get_mean_var(spd)
+    
+    def rollof(self, y):
+        """
+        Extracts the spectral rolloff of an audio signal.
+        
+        Parameters:
+            y (ndarray): The audio signal represented as an numpy array.
+        
+        Returns:
+            tuple: A tuple containing the mean and variance of the spectral rolloff.
+        """
+        rollof = librosa.feature.spectral_rolloff(y=y, sr=self.sr)[0]
+        return self.get_mean_var(rollof)
+    
+    def rms(self, y):
+        """
+        Extracts the root mean square (RMS) of an audio signal.
+        
+        Parameters:
+            y (ndarray): The audio signal represented as an numpy array.
+        
+        Returns:
+            tuple: A tuple containing the mean and variance of the RMS.
+        """
+        rms = librosa.feature.rms(y=y)
+        return self.get_mean_var(rms)
+    
+    def features(self,y):
+        """
+        Extracts various audio features from an audio signal.
+        
+        Parameters:
+            y (ndarray): The audio signal represented as an numpy array.
+        
+        Returns:
+            ndarray: An array containing the extracted audio features.
+        """
+        tempo = self.tempo(y)
+        centroid_mean, centroid_var = self.centroid(y)
+        chroma_mean, chroma_var = self.chroma_stft(y)
+        zcr_mean, zcr_var = self.zero_crossing_rate(y)
+        spd_mean, spd_var = self.spectral_bandwidth(y)
+        rollof_mean, rollof_var = self.rollof(y)
+        rsm_mean, rsm_var = self.rms(y)
+        harm, perc = self.harmonic_and_per(y)
+        harm_mean, harm_var = harm
+        perc_mean, perc_var = perc
+        mfccs = self.mfccs(y)
+        
+        features = np.array([y.shape[0],
+                            chroma_mean, chroma_var,
+                            rsm_mean, rsm_var,
+                            centroid_mean, centroid_var ,
+                            spd_mean, spd_var,
+                            rollof_mean, rollof_var,
+                            zcr_mean, zcr_var,
+                            harm_mean, harm_var,
+                            perc_mean, perc_var ,
+                            tempo,
+                           ],
+                 dtype=np.float32)
+        features = np.concatenate([features, mfccs])
+        return features
+    
+    def splits_3sec(self):
+        """
+        Splits an audio signal into 3-second sub-sequences and extracts audio features from each sub-sequence.
+        
+        Returns:
+            ndarray: An array containing the extracted audio features for each 3-second sub-sequence.
+        """
+        features_split = []
+        for sub_sequence in self.y:
+            feature = self.features(sub_sequence)
+            features_split.append(feature)
+        
+        features_np = np.array(features_split)
+        return features_np
+
+
+def load_model():
+    path =  os.path.dirname(__file__)
+    path_model = os.path.join(path, 'models', "model.pkl")
+    model = load(path_model)
+    return model
+
+def predict(features):
+    model = load_model()
+    prediction = model.predict(features)
+    mode = st.mode(prediction)
+    return CLASSES[mode], prediction
+
+def cuts_silence(audio):
+    audio_file, _ = librosa.effects.trim(audio)
+    return audio_file
+
+def convert_mp3_to_wav(music_file):  
+    name_file = "music_file.wav"
+    sound = AudioSegment.from_mp3(music_file)
+    sound.export(name_file,format="wav")
+    return name_file
+
+
+def preprosecing(uploaded_file):
+    name_file = convert_mp3_to_wav(uploaded_file)
+    y, sr = librosa.load(name_file)
+    audio_file = cuts_silence(y)
+    audio_file = audio_file[:sr*30]
+    sf.write(file=name_file, data=audio_file, samplerate=sr)
+    file = open(name_file, 'rb')
+    features = Features(audio_file, sr).splits_3sec()
+    prediction = predict(features)
+    return file, prediction