| Item type |
SIG Technical Reports(1) |
| 公開日 |
2016-08-29 |
| タイトル |
|
|
タイトル |
Studying mutual context of grasp types and object attributes in hand manipulation activities |
| タイトル |
|
|
言語 |
en |
|
タイトル |
Studying mutual context of grasp types and object attributes in hand manipulation activities |
| 言語 |
|
|
言語 |
eng |
| 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 |
|
|
|
The University of Tokyo |
| 著者所属 |
|
|
|
Carnegie Mellon University |
| 著者所属 |
|
|
|
The University of Tokyo |
| 著者所属(英) |
|
|
|
en |
|
|
The University of Tokyo |
| 著者所属(英) |
|
|
|
en |
|
|
Carnegie Mellon University |
| 著者所属(英) |
|
|
|
en |
|
|
The University of Tokyo |
| 著者名 |
Minjie, Cai
Kris, Kitani
Yoichi, Sato
|
| 著者名(英) |
Minjie, Cai
Kris, Kitani
Yoichi, Sato
|
| 論文抄録 |
|
|
内容記述タイプ |
Other |
|
内容記述 |
Recognizing hand grasp types and object attributes are important in understanding human hand manipulation activities. However, the recognition tasks are challenging as hands and objects are often occluded by each other during interactions. We observe that there is strong correlation between grasp types and object attributes, which can serve as mutual context in recognition. In this paper, we propose an unified model in which contextual relationship between grasp types and object attributes are studied. First, we develop a novel method to extract probabilistic information about grasp types and object attributes from still images in which hand-object interactions are recorded. We then explore the contextual relationship between grasp types and object attributes and show how the context information is used to boost the recognition of both. On public cooking datasets involving egocentric hand manipulation activities, experiment results strongly support our proposal. |
| 論文抄録(英) |
|
|
内容記述タイプ |
Other |
|
内容記述 |
Recognizing hand grasp types and object attributes are important in understanding human hand manipulation activities. However, the recognition tasks are challenging as hands and objects are often occluded by each other during interactions. We observe that there is strong correlation between grasp types and object attributes, which can serve as mutual context in recognition. In this paper, we propose an unified model in which contextual relationship between grasp types and object attributes are studied. First, we develop a novel method to extract probabilistic information about grasp types and object attributes from still images in which hand-object interactions are recorded. We then explore the contextual relationship between grasp types and object attributes and show how the context information is used to boost the recognition of both. On public cooking datasets involving egocentric hand manipulation activities, experiment results strongly support our proposal. |
| 書誌レコードID |
|
|
収録物識別子タイプ |
NCID |
|
収録物識別子 |
AA11131797 |
| 書誌情報 |
研究報告コンピュータビジョンとイメージメディア（CVIM）
巻 2016-CVIM-203,
号 20,
p. 1-8,
発行日 2016-08-29
|
| ISSN |
|
|
収録物識別子タイプ |
ISSN |
|
収録物識別子 |
2188-8701 |
| 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-CVIM16203020.pdf (1.4 MB)
