In this task, I have trained and evaluated Support Vector Machine (SVM) for prediction of genre. Essentia's MusicExtractor has been used to compute summarized descriptor values (no frames). Scikit-learn has been used for SVM classifier. All the features have been pre-processed first, by removing features with non-numerical irrelevant values and standardizing them to zero mean and unit variance.
This notebook along with the supporting .py files is self-sufficient in the sense that it handles everything from downloading the dataset to arriving at the classification results with appropriate instructions and explanations. For questions/feedback, please write to venkatesh.shenoykadandale01@estudiant.upf.edu.