Spatial–Temporal Knowledge-Embedded Transformer for Video Scene Graph Generation.
Saved in:
| Title: | Spatial–Temporal Knowledge-Embedded Transformer for Video Scene Graph Generation. |
|---|---|
| Authors: | Pu, Tao1 (AUTHOR) putao3@mail2.sysu.edu.cn, Chen, Tianshui2 (AUTHOR) tianshuichen@gmail.com, Wu, Hefeng1 (AUTHOR) wuhefeng@mail.sysu.edu.cn, Lu, Yongyi2 (AUTHOR) yylu1989@gmail.com, Lin, Liang1 (AUTHOR) linliang@ieee.org |
| Source: | IEEE Transactions on Image Processing. 2024, Vol. 33, p556-568. 13p. |
| Subjects: | Knowledge representation (Information theory), Statistical correlation, Videos, Prior learning |
| Abstract: | Video scene graph generation (VidSGG) aims to identify objects in visual scenes and infer their relationships for a given video. It requires not only a comprehensive understanding of each object scattered on the whole scene but also a deep dive into their temporal motions and interactions. Inherently, object pairs and their relationships enjoy spatial co-occurrence correlations within each image and temporal consistency/transition correlations across different images, which can serve as prior knowledge to facilitate VidSGG model learning and inference. In this work, we propose a spatial-temporal knowledge-embedded transformer (STKET) that incorporates the prior spatial-temporal knowledge into the multi-head cross-attention mechanism to learn more representative relationship representations. Specifically, we first learn spatial co-occurrence and temporal transition correlations in a statistical manner. Then, we design spatial and temporal knowledge-embedded layers that introduce the multi-head cross-attention mechanism to fully explore the interaction between visual representation and the knowledge to generate spatial- and temporal-embedded representations, respectively. Finally, we aggregate these representations for each subject-object pair to predict the final semantic labels and their relationships. Extensive experiments show that STKET outperforms current competing algorithms by a large margin, e.g., improving the mR@50 by 8.1%, 4.7%, and 2.1% on different settings over current algorithms. [ABSTRACT FROM AUTHOR] |
| Copyright of IEEE Transactions on Image Processing is the property of IEEE and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) | |
| Database: | Engineering Source |
| FullText | Text: Availability: 0 |
|---|---|
| Header | DbId: egs DbLabel: Engineering Source An: 174718021 AccessLevel: 6 PubType: Academic Journal PubTypeId: academicJournal PreciseRelevancyScore: 0 |
| IllustrationInfo | |
| Items | – Name: Title Label: Title Group: Ti Data: Spatial–Temporal Knowledge-Embedded Transformer for Video Scene Graph Generation. – Name: Author Label: Authors Group: Au Data: <searchLink fieldCode="AR" term="%22Pu%2C+Tao%22">Pu, Tao</searchLink><relatesTo>1</relatesTo> (AUTHOR)<i> putao3@mail2.sysu.edu.cn</i><br /><searchLink fieldCode="AR" term="%22Chen%2C+Tianshui%22">Chen, Tianshui</searchLink><relatesTo>2</relatesTo> (AUTHOR)<i> tianshuichen@gmail.com</i><br /><searchLink fieldCode="AR" term="%22Wu%2C+Hefeng%22">Wu, Hefeng</searchLink><relatesTo>1</relatesTo> (AUTHOR)<i> wuhefeng@mail.sysu.edu.cn</i><br /><searchLink fieldCode="AR" term="%22Lu%2C+Yongyi%22">Lu, Yongyi</searchLink><relatesTo>2</relatesTo> (AUTHOR)<i> yylu1989@gmail.com</i><br /><searchLink fieldCode="AR" term="%22Lin%2C+Liang%22">Lin, Liang</searchLink><relatesTo>1</relatesTo> (AUTHOR)<i> linliang@ieee.org</i> – Name: TitleSource Label: Source Group: Src Data: <searchLink fieldCode="JN" term="%22IEEE+Transactions+on+Image+Processing%22">IEEE Transactions on Image Processing</searchLink>. 2024, Vol. 33, p556-568. 13p. – Name: Subject Label: Subjects Group: Su Data: <searchLink fieldCode="DE" term="%22Knowledge+representation+%28Information+theory%29%22">Knowledge representation (Information theory)</searchLink><br /><searchLink fieldCode="DE" term="%22Statistical+correlation%22">Statistical correlation</searchLink><br /><searchLink fieldCode="DE" term="%22Videos%22">Videos</searchLink><br /><searchLink fieldCode="DE" term="%22Prior+learning%22">Prior learning</searchLink> – Name: Abstract Label: Abstract Group: Ab Data: Video scene graph generation (VidSGG) aims to identify objects in visual scenes and infer their relationships for a given video. It requires not only a comprehensive understanding of each object scattered on the whole scene but also a deep dive into their temporal motions and interactions. Inherently, object pairs and their relationships enjoy spatial co-occurrence correlations within each image and temporal consistency/transition correlations across different images, which can serve as prior knowledge to facilitate VidSGG model learning and inference. In this work, we propose a spatial-temporal knowledge-embedded transformer (STKET) that incorporates the prior spatial-temporal knowledge into the multi-head cross-attention mechanism to learn more representative relationship representations. Specifically, we first learn spatial co-occurrence and temporal transition correlations in a statistical manner. Then, we design spatial and temporal knowledge-embedded layers that introduce the multi-head cross-attention mechanism to fully explore the interaction between visual representation and the knowledge to generate spatial- and temporal-embedded representations, respectively. Finally, we aggregate these representations for each subject-object pair to predict the final semantic labels and their relationships. Extensive experiments show that STKET outperforms current competing algorithms by a large margin, e.g., improving the mR@50 by 8.1%, 4.7%, and 2.1% on different settings over current algorithms. [ABSTRACT FROM AUTHOR] – Name: AbstractSuppliedCopyright Label: Group: Ab Data: <i>Copyright of IEEE Transactions on Image Processing is the property of IEEE and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract.</i> (Copyright applies to all Abstracts.) |
| PLink | https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=egs&AN=174718021 |
| RecordInfo | BibRecord: BibEntity: Identifiers: – Type: doi Value: 10.1109/TIP.2023.3345652 Languages: – Code: eng Text: English PhysicalDescription: Pagination: PageCount: 13 StartPage: 556 Subjects: – SubjectFull: Knowledge representation (Information theory) Type: general – SubjectFull: Statistical correlation Type: general – SubjectFull: Videos Type: general – SubjectFull: Prior learning Type: general Titles: – TitleFull: Spatial–Temporal Knowledge-Embedded Transformer for Video Scene Graph Generation. Type: main BibRelationships: HasContributorRelationships: – PersonEntity: Name: NameFull: Pu, Tao – PersonEntity: Name: NameFull: Chen, Tianshui – PersonEntity: Name: NameFull: Wu, Hefeng – PersonEntity: Name: NameFull: Lu, Yongyi – PersonEntity: Name: NameFull: Lin, Liang IsPartOfRelationships: – BibEntity: Dates: – D: 01 M: 01 Text: 2024 Type: published Y: 2024 Identifiers: – Type: issn-print Value: 10577149 Numbering: – Type: volume Value: 33 Titles: – TitleFull: IEEE Transactions on Image Processing Type: main |
| ResultId | 1 |
