Spatio-temporal attention models for grounded video captioning
(2017) In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 10114 LNCS. p.104-119- Abstract
Automatic video captioning is challenging due to the complex interactions in dynamic real scenes. A comprehensive system would ultimately localize and track the objects, actions and interactions present in a video and generate a description that relies on temporal localization in order to ground the visual concepts. However, most existing automatic video captioning systems map from raw video data to high level textual description, bypassing localization and recognition, thus discarding potentially valuable information for content localization and generalization. In this work we present an automatic video captioning model that combines spatio-temporal attention and image classification by means of deep neural network structures based on... (More)
Automatic video captioning is challenging due to the complex interactions in dynamic real scenes. A comprehensive system would ultimately localize and track the objects, actions and interactions present in a video and generate a description that relies on temporal localization in order to ground the visual concepts. However, most existing automatic video captioning systems map from raw video data to high level textual description, bypassing localization and recognition, thus discarding potentially valuable information for content localization and generalization. In this work we present an automatic video captioning model that combines spatio-temporal attention and image classification by means of deep neural network structures based on long short-term memory. The resulting system is demonstrated to produce state-of-the-art results in the standard YouTube captioning benchmark while also offering the advantage of localizing the visual concepts (subjects, verbs, objects), with no grounding supervision, over space and time.
(Less)
- author
- Zanfir, Mihai ; Marinoiu, Elisabeta and Sminchisescu, Cristian LU
- organization
- publishing date
- 2017
- type
- Chapter in Book/Report/Conference proceeding
- publication status
- published
- subject
- host publication
- Computer Vision - 13th Asian Conference on Computer Vision, ACCV 2016, Revised Selected Papers
- series title
- Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
- volume
- 10114 LNCS
- pages
- 16 pages
- publisher
- Springer
- external identifiers
-
- scopus:85016047898
- ISSN
- 03029743
- 16113349
- ISBN
- 9783319541891
- DOI
- 10.1007/978-3-319-54190-7_7
- language
- English
- LU publication?
- yes
- id
- bddc74a2-9f6d-4215-8826-0cf95609c23f
- date added to LUP
- 2017-04-06 15:12:25
- date last changed
- 2025-01-20 12:30:06
@inbook{bddc74a2-9f6d-4215-8826-0cf95609c23f, abstract = {{<p>Automatic video captioning is challenging due to the complex interactions in dynamic real scenes. A comprehensive system would ultimately localize and track the objects, actions and interactions present in a video and generate a description that relies on temporal localization in order to ground the visual concepts. However, most existing automatic video captioning systems map from raw video data to high level textual description, bypassing localization and recognition, thus discarding potentially valuable information for content localization and generalization. In this work we present an automatic video captioning model that combines spatio-temporal attention and image classification by means of deep neural network structures based on long short-term memory. The resulting system is demonstrated to produce state-of-the-art results in the standard YouTube captioning benchmark while also offering the advantage of localizing the visual concepts (subjects, verbs, objects), with no grounding supervision, over space and time.</p>}}, author = {{Zanfir, Mihai and Marinoiu, Elisabeta and Sminchisescu, Cristian}}, booktitle = {{Computer Vision - 13th Asian Conference on Computer Vision, ACCV 2016, Revised Selected Papers}}, isbn = {{9783319541891}}, issn = {{03029743}}, language = {{eng}}, pages = {{104--119}}, publisher = {{Springer}}, series = {{Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)}}, title = {{Spatio-temporal attention models for grounded video captioning}}, url = {{http://dx.doi.org/10.1007/978-3-319-54190-7_7}}, doi = {{10.1007/978-3-319-54190-7_7}}, volume = {{10114 LNCS}}, year = {{2017}}, }