Final project of KTH course DT2119 Speech and Speaker Recognition. This project consists of employing different approaches in order to perform music genre classification over the FMA dataset.
The experiments include from the simplest approaches such as SVM and K-NN to deep learning approaches by means of different configurations of convolutional neural networks, which can be found in Models.
To run the experiments:
- Go to base_models to run the simple approaches.
- Go to final_models to run the deep learning approaches.
- Modify the paths in constants if needed (as it is in order to run in Colab).
- Modify the network parameters to customize the architecture and performance.
- Run the selected code (it needs TensorFlow in its version 2 to run).
The conclsusions of the different experiments can be found in MusicGenreRecognition_Group4.