Details
| Originalsprache | Englisch |
|---|---|
| Titel des Sammelwerks | Computer Vision Systems |
| Untertitel | 8th International Conference, ICVS 2011, Proceedings |
| Seiten | 71-80 |
| Seitenumfang | 10 |
| Publikationsstatus | Veröffentlicht - 2011 |
| Extern publiziert | Ja |
| Veranstaltung | 8th International Conference on Computer Vision Systems, ICVS 2011 - Sophia Antipolis, Frankreich Dauer: 20 Sept. 2011 → 22 Sept. 2011 |
Publikationsreihe
| Name | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) |
|---|---|
| Band | 6962 LNCS |
| ISSN (Print) | 0302-9743 |
| ISSN (elektronisch) | 1611-3349 |
Abstract
State-of-the-art systems for visual concept detection typically rely on the Bag-of-Visual-Words representation. While several aspects of this representation have , such as keypoint sampling strategy, vocabulary size, projection method, weighting scheme or the integration of color, the impact of the spatial extents of local SIFT descriptors has not been studied in previous work. In this paper, the effect of different spatial extents in a state-of-the-art system for visual concept detection is investigated. Based on the observation that SIFT descriptors with different spatial extents yield large performance differences, we propose a concept detection system that combines feature representations for different spatial extents using multiple kernel learning. It is shown experimentally on a large set of 101 concepts from the Mediamill Challenge and on the PASCAL Visual Object Classes Challenge that these feature representations are complementary: Superior performance can be achieved on both test sets using the proposed system.
ASJC Scopus Sachgebiete
- Mathematik (insg.)
- Theoretische Informatik
- Informatik (insg.)
- Allgemeine Computerwissenschaft
Zitieren
- Standard
- Harvard
- Apa
- Vancouver
- BibTex
- RIS
Computer Vision Systems : 8th International Conference, ICVS 2011, Proceedings. 2011. S. 71-80 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Band 6962 LNCS).
Publikation: Beitrag in Buch/Bericht/Sammelwerk/Konferenzband › Aufsatz in Konferenzband › Forschung › Peer-Review
}
TY - GEN
T1 - On the spatial extents of SIFT descriptors for visual concept detection
AU - Mühling, Markus
AU - Ewerth, Ralph
AU - Freisleben, Bernd
PY - 2011
Y1 - 2011
N2 - State-of-the-art systems for visual concept detection typically rely on the Bag-of-Visual-Words representation. While several aspects of this representation have , such as keypoint sampling strategy, vocabulary size, projection method, weighting scheme or the integration of color, the impact of the spatial extents of local SIFT descriptors has not been studied in previous work. In this paper, the effect of different spatial extents in a state-of-the-art system for visual concept detection is investigated. Based on the observation that SIFT descriptors with different spatial extents yield large performance differences, we propose a concept detection system that combines feature representations for different spatial extents using multiple kernel learning. It is shown experimentally on a large set of 101 concepts from the Mediamill Challenge and on the PASCAL Visual Object Classes Challenge that these feature representations are complementary: Superior performance can be achieved on both test sets using the proposed system.
AB - State-of-the-art systems for visual concept detection typically rely on the Bag-of-Visual-Words representation. While several aspects of this representation have , such as keypoint sampling strategy, vocabulary size, projection method, weighting scheme or the integration of color, the impact of the spatial extents of local SIFT descriptors has not been studied in previous work. In this paper, the effect of different spatial extents in a state-of-the-art system for visual concept detection is investigated. Based on the observation that SIFT descriptors with different spatial extents yield large performance differences, we propose a concept detection system that combines feature representations for different spatial extents using multiple kernel learning. It is shown experimentally on a large set of 101 concepts from the Mediamill Challenge and on the PASCAL Visual Object Classes Challenge that these feature representations are complementary: Superior performance can be achieved on both test sets using the proposed system.
KW - Bag-of-Words
KW - Magnification Factor
KW - SIFT
KW - Spatial Bin Size
KW - Video Retrieval
KW - Visual Concept Detection
UR - http://www.scopus.com/inward/record.url?scp=80053457242&partnerID=8YFLogxK
U2 - 10.1007/978-3-642-23968-7_8
DO - 10.1007/978-3-642-23968-7_8
M3 - Conference contribution
AN - SCOPUS:80053457242
SN - 9783642239670
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 71
EP - 80
BT - Computer Vision Systems
T2 - 8th International Conference on Computer Vision Systems, ICVS 2011
Y2 - 20 September 2011 through 22 September 2011
ER -