| Item type |
SIG Technical Reports(1) |
| 公開日 |
2024-05-08 |
| タイトル |
|
|
タイトル |
Mobile Sensor-Based Indoor Object Searching with Visual-Language Model |
| タイトル |
|
|
言語 |
en |
|
タイトル |
Mobile Sensor-Based Indoor Object Searching with Visual-Language Model |
| 言語 |
|
|
言語 |
eng |
| キーワード |
|
|
主題Scheme |
Other |
|
主題 |
[MBL]センシングとIoT |
| 資源タイプ |
|
|
資源タイプ識別子 |
http://purl.org/coar/resource_type/c_18gh |
|
資源タイプ |
technical report |
| 著者所属 |
|
|
|
大阪大学大学院情報科学研究科 |
| 著者所属 |
|
|
|
大阪大学大学院情報科学研究科 |
| 著者所属 |
|
|
|
大阪大学大学院情報科学研究科 |
| 著者名 |
Haruki, Yonekura
Hamada, Rizk
Hirozumi, Yamaguchi
|
| 著者名(英) |
Haruki, Yonekura
Hamada, Rizk
Hirozumi, Yamaguchi
|
| 論文抄録 |
|
|
内容記述タイプ |
Other |
|
内容記述 |
In this study, we tackle the challenge of automatically identifying and classifying items in indoor environments. Traditional methods struggle with this application due to the high cost of labeling items and the inherent ambiguity of natural language descriptions. To address this, we present a novel object-searching system that leverages a visual-language model. This system is trained on rich indoor environmental data, including detailed point clouds and RGB images, which can be currently collected by sensors installed in a mobile phone, enabling users to adapt to changes in the indoor environment. It allows users to search for items using natural language, eliminating the need for pre-defined labels. The effectiveness of the system is evaluated through experiments that analyze performance metrics like accuracy and efficiency with realistic datasets, demonstrating its practical usefulness. |
| 論文抄録(英) |
|
|
内容記述タイプ |
Other |
|
内容記述 |
In this study, we tackle the challenge of automatically identifying and classifying items in indoor environments. Traditional methods struggle with this application due to the high cost of labeling items and the inherent ambiguity of natural language descriptions. To address this, we present a novel object-searching system that leverages a visual-language model. This system is trained on rich indoor environmental data, including detailed point clouds and RGB images, which can be currently collected by sensors installed in a mobile phone, enabling users to adapt to changes in the indoor environment. It allows users to search for items using natural language, eliminating the need for pre-defined labels. The effectiveness of the system is evaluated through experiments that analyze performance metrics like accuracy and efficiency with realistic datasets, demonstrating its practical usefulness. |
| 書誌レコードID |
|
|
収録物識別子タイプ |
NCID |
|
収録物識別子 |
AA11851388 |
| 書誌情報 |
研究報告モバイルコンピューティングと新社会システム（MBL）
巻 2024-MBL-111,
号 42,
p. 1-7,
発行日 2024-05-08
|
| ISSN |
|
|
収録物識別子タイプ |
ISSN |
|
収録物識別子 |
2188-8817 |
| Notice |
|
|
|
SIG Technical Reports are nonrefereed and hence may later appear in any journals, conferences, symposia, etc. |
| 出版者 |
|
|
言語 |
ja |
|
出版者 |
情報処理学会 |
IPSJ-MBL24111042.pdf (7.7 MB)
