| IEICE Technical Committee Submission System Conference Paper's Information |
Online Proceedings [Sign in] Tech. Rep. Archives |
| Go Top Page | Go Previous | [Japanese] / [English] |
| Paper Abstract and Keywords | ||
| Presentation | 2021-06-19 15:00 Investigation on fine-tuning with image classification networks for deep neural network-based musical instrument classification Yuki Shiroma, Yuma Kinoshita, Sayaka Shiota, Hitoshi Kiya (TMU)  SP2021-17 |
|
| Abstract | (in Japanese) | (See Japanese page) |
| (in English) | In this paper, we investigate abilities of channel conversion methods for fine-tuning with image classification networks under deep neural network-based musical instrument classification. Recently, many deep neural network-based methods have been proposed for scene classification, emotion recognition tasks, and so on. It has also been reported that fine-tuning techniques with well-trained networks using large-scale image dataset improve the performance of sound classification tasks when the limited amount of training data is available. In this case, while a spectrogram extracted from sound data is usually regarded as an image and inputted to the fine-tuned networks with the image classification tasks, the spectrogram image is not suitable to the fine-tuned network because the input of the image classification networks assumes the three channel data like RGB. In this case, the spectrogram is required to be converted to the three channel data, and many methods such as spectrogram duplication method, a method using delta as coefficients and colorization of a spectrogram have been proposed. However, there is no discussion how these methods affect the accuracies. Therefore, we compare various channel conversion methods via fine-tuning of the image classification networks. In the experiments, we performed musical instrument classificaiton with fine-tuning of the well-trained networks by ImageNet. From the results, compared among six channel conversion methods, the colorization of a spectrogram was the most suitable for the fine-tuning with the image classification networks. | |
| Keyword | (in Japanese) | (See Japanese page) |
| (in English) | Acoustic musical instrument classification / image classification network / fine-tuning / channel conversion / / / / | |
| Reference Info. | IEICE Tech. Rep., vol. 121, no. 66, SP2021-17, pp. 75-79, June 2021. | |
| Paper # | SP2021-17 | |
| Date of Issue | 2021-06-11 (SP) | |
| ISSN | Online edition: ISSN 2432-6380 | |
| Copyright and reproduction |
All rights are reserved and no part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Notwithstanding, instructors are permitted to photocopy isolated articles for noncommercial classroom use without fee. (License No.: 10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) | |
| Download PDF | SP2021-17 |
|
| Conference Information | |
| Committee | SP IPSJ-SLP IPSJ-MUS |
| Conference Date | 2021-06-18 - 2021-06-19 |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | Online |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | OTOGAKU Symposium 2021 |
| Paper Information | |
| Registration To | SP |
| Conference Code | 2021-06-SP-SLP-MUS |
| Language | Japanese |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Investigation on fine-tuning with image classification networks for deep neural network-based musical instrument classification |
| Sub Title (in English) | |
| Keyword(1) | Acoustic musical instrument classification |
| Keyword(2) | image classification network |
| Keyword(3) | fine-tuning |
| Keyword(4) | channel conversion |
| Keyword(5) | |
| Keyword(6) | |
| Keyword(7) | |
| Keyword(8) | |
| 1st Author's Name | Yuki Shiroma |
| 1st Author's Affiliation | Tokyo Metropolitan University (TMU) |
| 2nd Author's Name | Yuma Kinoshita |
| 2nd Author's Affiliation | Tokyo Metropolitan University (TMU) |
| 3rd Author's Name | Sayaka Shiota |
| 3rd Author's Affiliation | Tokyo Metropolitan University (TMU) |
| 4th Author's Name | Hitoshi Kiya |
| 4th Author's Affiliation | Tokyo Metropolitan University (TMU) |
| 5th Author's Name | |
| 5th Author's Affiliation | () |
| 6th Author's Name | |
| 6th Author's Affiliation | () |
| 7th Author's Name | |
| 7th Author's Affiliation | () |
| 8th Author's Name | |
| 8th Author's Affiliation | () |
| 9th Author's Name | |
| 9th Author's Affiliation | () |
| 10th Author's Name | |
| 10th Author's Affiliation | () |
| 11th Author's Name | |
| 11th Author's Affiliation | () |
| 12th Author's Name | |
| 12th Author's Affiliation | () |
| 13th Author's Name | |
| 13th Author's Affiliation | () |
| 14th Author's Name | |
| 14th Author's Affiliation | () |
| 15th Author's Name | |
| 15th Author's Affiliation | () |
| 16th Author's Name | |
| 16th Author's Affiliation | () |
| 17th Author's Name | |
| 17th Author's Affiliation | () |
| 18th Author's Name | |
| 18th Author's Affiliation | () |
| Speaker | Author-1 |
| Date Time | 2021-06-19 15:00:00 |
| Presentation Time | 120 minutes |
| Registration for | SP |
| Paper # | SP2021-17 |
| Volume (vol) | vol.121 |
| Number (no) | no.66 |
| Page | pp.75-79 |
| #Pages | 5 |
| Date of Issue | 2021-06-11 (SP) |
[Return to Top Page]