Implementing different machine learning models on input music data and finding the most efficient method.
We have compared KNN and other ensemble features on GTZAN Dataset
*model.py
We Scale our features using scaler()
Training of model on logisitic regression
We select main 30 audio features out of the given 55 using feature permutance
We now train those models on various
Hyperparameter tuning is performed
Those trained weights are saved in pickle files
*app.py
Flask is imported
The program extracts input from user in index.html
The input saved in file extracts audio data using librosa library
Feature extraction takes place (the same important 30 features are extracted)
Refer the official librosa docs for this
We load the pickle files of trained models and run on the data
The input is splitted into 10 intervals of 3 seconds out of 30 seconds audio as to compare to the dataset where similar has been done
It is broken to create a much bigger dataset from much lesser data
We test the data input data and send the results back to index.html
*index.html
Front-end for the project which extracts the audio file for user
plays it
Returns the recognized genre and model used along\
You can find the hosted website here: Click to View
Check my published paper on this topic Music Genre Classification Using Machine Learning Models
