Details
| Originalsprache | Englisch |
|---|---|
| Titel des Sammelwerks | ICMR 2016 - Proceedings of the 2016 ACM International Conference on Multimedia Retrieval |
| Seiten | 377-380 |
| Seitenumfang | 4 |
| ISBN (elektronisch) | 9781450343596 |
| Publikationsstatus | Veröffentlicht - 6 Juni 2016 |
| Veranstaltung | 6th ACM International Conference on Multimedia Retrieval, ICMR 2016 - New York, USA / Vereinigte Staaten Dauer: 6 Juni 2016 → 9 Juni 2016 |
Publikationsreihe
| Name | ICMR 2016 - Proceedings of the 2016 ACM International Conference on Multimedia Retrieval |
|---|
Abstract
Deep neural networks have been successfully applied to the task of visual concept classification. However, they require a large number of training examples for learning. Although pre-trained deep neural networks are available for some domains, they usually have to be fine-tuned for an envisaged target domain. Recently, some approaches have been suggested that are aimed at incrementally (or even endlessly) learning visual concepts based on Web data. Since tags of Web images are often noisy, normally some filtering mechanisms are employed in order to remove "spam" images that are not appropriate for training. In this paper, we investigate several aspects of a web-supervised system that has to be adapted to another target domain: 1.) the effect of incremental learning, 2.) the effect of spam filtering, and 3.) the behavior of particular concept classes with respect to 1.) and 2.). The experimental results provide some insights under which conditions incremental learning and spam filtering are useful.
ASJC Scopus Sachgebiete
- Informatik (insg.)
- Computergrafik und computergestütztes Design
- Informatik (insg.)
- Mensch-Maschine-Interaktion
- Informatik (insg.)
- Software
Zitieren
- Standard
- Harvard
- Apa
- Vancouver
- BibTex
- RIS
ICMR 2016 - Proceedings of the 2016 ACM International Conference on Multimedia Retrieval. 2016. S. 377-380 (ICMR 2016 - Proceedings of the 2016 ACM International Conference on Multimedia Retrieval).
Publikation: Beitrag in Buch/Bericht/Sammelwerk/Konferenzband › Aufsatz in Konferenzband › Forschung › Peer-Review
}
TY - GEN
T1 - On the effects of spam filtering and incremental learning for web-supervised visual concept classification
AU - Springstein, Matthias
AU - Ewerth, Ralph
PY - 2016/6/6
Y1 - 2016/6/6
N2 - Deep neural networks have been successfully applied to the task of visual concept classification. However, they require a large number of training examples for learning. Although pre-trained deep neural networks are available for some domains, they usually have to be fine-tuned for an envisaged target domain. Recently, some approaches have been suggested that are aimed at incrementally (or even endlessly) learning visual concepts based on Web data. Since tags of Web images are often noisy, normally some filtering mechanisms are employed in order to remove "spam" images that are not appropriate for training. In this paper, we investigate several aspects of a web-supervised system that has to be adapted to another target domain: 1.) the effect of incremental learning, 2.) the effect of spam filtering, and 3.) the behavior of particular concept classes with respect to 1.) and 2.). The experimental results provide some insights under which conditions incremental learning and spam filtering are useful.
AB - Deep neural networks have been successfully applied to the task of visual concept classification. However, they require a large number of training examples for learning. Although pre-trained deep neural networks are available for some domains, they usually have to be fine-tuned for an envisaged target domain. Recently, some approaches have been suggested that are aimed at incrementally (or even endlessly) learning visual concepts based on Web data. Since tags of Web images are often noisy, normally some filtering mechanisms are employed in order to remove "spam" images that are not appropriate for training. In this paper, we investigate several aspects of a web-supervised system that has to be adapted to another target domain: 1.) the effect of incremental learning, 2.) the effect of spam filtering, and 3.) the behavior of particular concept classes with respect to 1.) and 2.). The experimental results provide some insights under which conditions incremental learning and spam filtering are useful.
KW - Deep convolutional neural network
KW - Visual concept classification
KW - Web-supervised learning
UR - http://www.scopus.com/inward/record.url?scp=84978761893&partnerID=8YFLogxK
U2 - 10.1145/2911996.2912072
DO - 10.1145/2911996.2912072
M3 - Conference contribution
AN - SCOPUS:84978761893
T3 - ICMR 2016 - Proceedings of the 2016 ACM International Conference on Multimedia Retrieval
SP - 377
EP - 380
BT - ICMR 2016 - Proceedings of the 2016 ACM International Conference on Multimedia Retrieval
T2 - 6th ACM International Conference on Multimedia Retrieval, ICMR 2016
Y2 - 6 June 2016 through 9 June 2016
ER -