ABSTRACT
The number of images uploaded to the web is enormous and is rapidly increasing. The purpose of our work is to use these for acquiring positive training data for visual concept learning. Manually creating training data for visual concept classifiers is an expensive and time consuming task. We propose an approach which automatically collects positive training samples from the Web by constructing a multitude of text queries and retaining for each query only very few top-ranked images returned by each one of the different web image search engines (Google, Flickr and Bing). In this way, we sift the burden of false positive rejection to the Web search engines and directly assemble a rich set of high-quality positive training samples. Experiments on forty concepts, evaluated on the ImageNet dataset, show the merit of the proposed approach.
- P. Over, G. Awad, et al., "Trecvid 2014 - an overview of the goals, tasks, data, evaluation mechanisms and metrics," in Proc. TRECVID 2014. NIST, USA, 2014.Google Scholar
- X. Li, C. G. Snoek, et al., "Harvesting social images for bi-concept search," IEEE Trans. on Multimedia, vol. 14, no. 4, pp. 1091--1104, 2012. Google ScholarDigital Library
- X. Li, C. G. Snoek, and M. Worring, "Unsupervised multi-feature tag relevance learning for social image retrieval," in Proc. Int. Conf. on Image and Video Retrieval. ACM, 2010, pp. 10--17. Google ScholarDigital Library
- J. Deng, W. Dong, et al., "Imagenet: A large-scale hierarchical image database," in Proc. Int. Conf. on Computer Vision and Pattern Recognition. IEEE, 2009, pp. 248--255.Google Scholar
- S. Zhu, C.-W. Ngo, and Y.-G. Jiang, "Sampling and ontologically pooling web images for visual concept learning," IEEE Trans. on Multimedia, vol. 14, no. 4, pp. 1068--1078, 2012. Google ScholarDigital Library
- F. Schroff, A. Criminisi, and A. Zisserman, "Harvesting image databases from the web," IEEE Trans. on Pattern Analysis and Machine Intelligence, vol. 33, no. 4, pp. 754--766, 2011. Google ScholarDigital Library
- L.-J. Li and L. Fei-Fei, "Optimol: automatic online picture collection via incremental model learning," Int. Journal of Computer Vision, vol. 88, no. 2, pp. 147--168, 2010. Google ScholarDigital Library
- X. Li and C. G. Snoek, "Visual categorization with negative examples for free," in Proc. 17th ACM Int. Conf. on Multimedia, 2009, pp. 661--664. Google ScholarDigital Library
- C. Fellbaum, WordNet: An Electronic Lexical Database, Bradford Books, 1998.Google ScholarCross Ref
- L. Han, A. Kashyap, et al., "Umbc ebiquity-core: Semantic textual similarity systems," in Proc. 2nd Joint Conf. on Lexical and Computational Semantics, 2013, vol. 1, pp. 44--52.Google Scholar
- D. Carvalho, C. Callı, et al., "Easyesa: A low-effort infrastructure for explicit semantic analysis," in Proc. 13th Int. Semantic Web Conference (ISWC), 2014.Google Scholar
- J. Ramos, "Using tf-idf to determine word relevance in document queries," in Proc. 1st Instructional Conf. on Machine Learning, 2003.Google Scholar
- F. Markatopoulou, N. Pittaras, et al., "A study on the use of a binary local descriptor and color extensions of local descriptors for video concept detection," in Proc. MultiMedia Modeling, Springer, 2015, pp. 282--293.Google Scholar
- A. F. Smeaton, P. Over, and W. Kraaij, "Evaluation campaigns and trecvid," in Proc. 8th ACM Int. Workshop on Multimedia Information Retrieval, 2006, pp. 321--330. Google ScholarDigital Library
Index Terms
- Exploiting Multiple Web Resources towards Collecting Positive Training Samples for Visual Concept Learning
Recommendations
Exploiting generalized discriminative multiple instance learning for multimedia semantic concept detection
A generalized discriminative multiple instance learning (GDMIL) algorithm is presented to train the classifier in the condition of vague annotation of training samples GDMIL not only inherits the original MIL's capability of automatically weighting the ...
Sampling and Ontologically Pooling Web Images for Visual Concept Learning
Part 1Sufficient training examples are essential for effective learning of semantic visual concepts. In practice, however, acquiring noise-free training examples has always been expensive. Recently the rapid popularization of social media websites, such as ...
On the pooling of positive examples with ontology for visual concept learning
MM '11: Proceedings of the 19th ACM international conference on MultimediaA common obstacle in effective learning of visual concept classifiers is the scarcity of positive training examples due to expensive labeling cost. This paper explores the sampling of weakly tagged web images for concept learning without human ...
Comments