Cool idea for visualization of audio tracks. It takes ~0.5s-2s of audio sliding window uses an [ ConvNet ] -based [ auto-encoder ] to compress it to 3 dimensions.
Representation of songs as trajectories in a 3D audio embedding
webvideoUrl
YT video
GH repo
Cool idea for visualization of audio tracks. It takes ~0.5s-2s of audio sliding window uses an [ ConvNet ] -based [ auto-encoder ] to compress it to 3 dimensions.