Presentation | 2022-09-09 An Image Recognition Model of Danger Objects for Diverse Clients using Federated Learning Yasuhiro Nitta, Ryo Yonetani, Maki Sugimoto, Hideo Saito, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | A disabled person can have cognition of danger objects during walking, which might not coincide with a non-disabled person. Also, even if we train a machine learning model with similar disabilities, the model should have the capability to accommodate the diverse cognition between person and person. Therefore, we must optimize the machine learning model for disabled people for each user. To get the capability, a large set of images taken under various users is required to train the optimized machine learning model. However, collecting such a large number of training data is challenging. Furthermore, from the privacy protection perspective, avoiding uploading personal images to the cloud is also desirable. Thus, to examine the capability in consideration of privacy protection, we construct a machine learning model of danger objects with diverse characteristics of users using federated learning with hierarchical clustering. As the result of the experiment, the constructed machine learning model showed higher accuracy than a conventional learning method, which was trained only on each client’s data. In addition, we analyzed the performance of the machine learning model by changing the number of clusters. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Federated Learning / Hierarchical Clustering / Image Recognition / Physical Disability Support |
Paper # | MVE2022-15 |
Date of Issue | 2022-09-01 (MVE) |
Conference Information | |
Committee | MVE |
---|---|
Conference Date | 2022/9/8(2days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | |
Chair | Kiyoshi Kiyokawa(NAIST) |
Vice Chair | Sumaru Niida(KDDI Research) |
Secretary | Sumaru Niida(NAIST) |
Assistant | Hidehiko Shishido(Univ. of Tsukuba) / Atsushi Nakazawa(Kyoto Univ.) / Naoya Tojo(KDDI Research) / Naoki Hagiyama(NTT) |
Paper Information | |
Registration To | Technical Committee on Media Experience and Virtual Environment |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | An Image Recognition Model of Danger Objects for Diverse Clients using Federated Learning |
Sub Title (in English) | |
Keyword(1) | Federated Learning |
Keyword(2) | Hierarchical Clustering |
Keyword(3) | Image Recognition |
Keyword(4) | Physical Disability Support |
1st Author's Name | Yasuhiro Nitta |
1st Author's Affiliation | Keio University(Keio Univ.) |
2nd Author's Name | Ryo Yonetani |
2nd Author's Affiliation | Keio University(Keio Univ.) |
3rd Author's Name | Maki Sugimoto |
3rd Author's Affiliation | Keio University(Keio Univ.) |
4th Author's Name | Hideo Saito |
4th Author's Affiliation | Keio University(Keio Univ.) |
Date | 2022-09-09 |
Paper # | MVE2022-15 |
Volume (vol) | vol.122 |
Number (no) | MVE-175 |
Page | pp.pp.26-31(MVE), |
#Pages | 6 |
Date of Issue | 2022-09-01 (MVE) |