Revisiting the onsets and frames model with additive attention
dc.contributor.author | Cheuk, KW | en_US |
dc.contributor.author | Luo, Y-J | en_US |
dc.contributor.author | Benetos, E | en_US |
dc.contributor.author | Herremans, D | en_US |
dc.contributor.author | International Joint Conference on Neural Networks (IJCNN) | en_US |
dc.date.accessioned | 2021-05-25T15:22:39Z | |
dc.date.available | 2021-04-10 | en_US |
dc.date.issued | 2021-07-18 | en_US |
dc.identifier.uri | https://qmro.qmul.ac.uk/xmlui/handle/123456789/72070 | |
dc.description.abstract | Recent advances in automatic music transcription (AMT) have achieved highly accurate polyphonic piano transcription results by incorporating onset and offset detection. The existing literature, however, focuses mainly on the leverage of deep and complex models to achieve state-of-the-art (SOTA) accuracy, without understanding model behaviour. In this paper, we conduct a comprehensive examination of the Onsets-and-Frames AMT model, and pinpoint the essential components contributing to a strong AMT performance. This is achieved through exploitation of a modified additive attention mechanism. The experimental results suggest that the attention mechanism beyond a moderate temporal context does not benefit the model, and that rule-based post-processing is largely responsible for the SOTA performance. We also demonstrate that the onsets are the most significant attentive feature regardless of model complexity. The findings encourage AMT research to weigh more on both a robust onset detector and an effective post-processor. | en_US |
dc.format.extent | ? - ? (8) | en_US |
dc.publisher | IEEE | en_US |
dc.relation.replaces | 123456789/71866 | |
dc.relation.replaces | https://qmro.qmul.ac.uk/xmlui/handle/123456789/71866 | |
dc.rights | © 2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. | * |
dc.rights.uri | http://creativecommons.org/licenses/by/3.0/us/ | * |
dc.title | Revisiting the onsets and frames model with additive attention | en_US |
dc.type | Conference Proceeding | |
pubs.merge-from | 123456789/71866 | |
pubs.merge-from | https://qmro.qmul.ac.uk/xmlui/handle/123456789/71866 | |
pubs.notes | Not known | en_US |
pubs.publication-status | Accepted | en_US |
pubs.publisher-url | https://www.ijcnn.org/ | en_US |
dcterms.dateAccepted | 2021-04-10 | en_US |
Files in this item
This item appears in the following Collection(s)
Except where otherwise noted, this item's license is described as © 2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.