Unsupervised Musical Object Discovery from Audio
This work addresses unsupervised object discovery in music audio, which is incremental as it adapts an existing visual method to a new domain.
The paper tackles unsupervised music decomposition by adapting SlotAttention to audio, introducing MusicSlots to address challenges like the lack of auditory occlusion analogues, and achieves good performance on note discovery and outperforms baselines on supervised prediction tasks.
Current object-centric learning models such as the popular SlotAttention architecture allow for unsupervised visual scene decomposition. Our novel MusicSlots method adapts SlotAttention to the audio domain, to achieve unsupervised music decomposition. Since concepts of opacity and occlusion in vision have no auditory analogues, the softmax normalization of alpha masks in the decoders of visual object-centric models is not well-suited for decomposing audio objects. MusicSlots overcomes this problem. We introduce a spectrogram-based multi-object music dataset tailored to evaluate object-centric learning on western tonal music. MusicSlots achieves good performance on unsupervised note discovery and outperforms several established baselines on supervised note property prediction tasks.