/MaSC_sim_vis

music similarity visualization with mfccs, pca, autoencoder, t-SNE | part of DLfM2019 submission

Primary LanguageJupyter Notebook

MaSC Compendium Visualization

A collection of scripts for visualizing the Arab Mashriq collection of the NYU Abu Dhabi Library and the Eisenberg collection

mfcc_t-SNE.ipynb: Compute MFCC from audio, reduce dimension to 2 with t-SNE, and plot

chromagram_t-SNE.ipynb: Compute chromagram from audio, reduce dimension to 3 and 2 with t-SNE, and plot

pca_t-SNE.ipynb: Compute mel spectrogram from audio, do PCA using different number of components, reduce dimension to 2, and plot alongside intensity

autoencoder_t_SNE: Compute mel spectrogram from audio, take bottleneck of autoencoder, reduce dimension to 2, and plot alongside intensity

compute_features.py: Compute, save, and plot STFT, chroma, and MFCC from audio

startAD_demo.ipynb: Independent similarity axis traversal visualization of MFCC and chroma from audio for startAD demo

features_VR.py: Compute selected features from audio, cross-refence song title with unique tag, and output csv with coordinates for VR