@InProceedings{	  toro:etal:2017,
  author	= {R. Toro, J. Baier, C. Ruz, A. Soto},
  title		= {How a General-Purpose Commonsense Ontology can Improve
		  Performance of Learning-Based Image Retrieval},
  booktitle	= {{IJCAI}},
  year		= {2017},
  abstract	= {The knowledge representation community has built
		  general-purpose ontologies which contain large amounts of
		  commonsense knowledge over relevant aspects of the world,
		  including useful visual information, e.g.: "a ball is used
		  by a football player", "a tennis player is located at a
		  tennis court". Current state-of-the-art approaches for
		  visual recognition do not exploit these rule-based
		  knowledge sources. Instead, they learn recognition models
		  directly from training examples. In this paper, we study
		  how general-purpose ontologies---specifically, MIT's
		  ConceptNet ontology---can improve the performance of
		  state-of-the-art vision systems. As a testbed, we tackle
		  the problem of sentence-based image retrieval. Our
		  retrieval approach incorporates knowledge from ConceptNet
		  on top of a large pool of object detectors derived from a
		  deep learning technique. In our experiments, we show that
		  ConceptNet can improve performance on a common benchmark
		  dataset. Key to our performance is the use of the ESPGAME
		  dataset to select visually relevant relations from
		  ConceptNet. Consequently, a main conclusion of this work is
		  that general-purpose commonsense ontologies improve
		  performance on visual reasoning tasks when properly
		  filtered to select meaningful visual relations. },
  url		= {https://arxiv.org/abs/1705.08844}
}

{"_id":"ufgKFkLcbeCbiqz8j","bibbaseid":"rtoro-cruz-howageneralpurposecommonsenseontologycanimproveperformanceoflearningbasedimageretrieval-2017","author_short":["R. Toro, J. B.","C. Ruz, A. S."],"bibdata":{"bibtype":"inproceedings","type":"inproceedings","author":[{"propositions":[],"lastnames":["R.","Toro"],"firstnames":["J.","Baier"],"suffixes":[]},{"propositions":[],"lastnames":["C.","Ruz"],"firstnames":["A.","Soto"],"suffixes":[]}],"title":"How a General-Purpose Commonsense Ontology can Improve Performance of Learning-Based Image Retrieval","booktitle":"IJCAI","year":"2017","abstract":"The knowledge representation community has built general-purpose ontologies which contain large amounts of commonsense knowledge over relevant aspects of the world, including useful visual information, e.g.: \"a ball is used by a football player\", \"a tennis player is located at a tennis court\". Current state-of-the-art approaches for visual recognition do not exploit these rule-based knowledge sources. Instead, they learn recognition models directly from training examples. In this paper, we study how general-purpose ontologies—specifically, MIT's ConceptNet ontology—can improve the performance of state-of-the-art vision systems. As a testbed, we tackle the problem of sentence-based image retrieval. Our retrieval approach incorporates knowledge from ConceptNet on top of a large pool of object detectors derived from a deep learning technique. In our experiments, we show that ConceptNet can improve performance on a common benchmark dataset. Key to our performance is the use of the ESPGAME dataset to select visually relevant relations from ConceptNet. Consequently, a main conclusion of this work is that general-purpose commonsense ontologies improve performance on visual reasoning tasks when properly filtered to select meaningful visual relations. ","url":"https://arxiv.org/abs/1705.08844","bibtex":"@InProceedings{\t toro:etal:2017,\n author\t= {R. Toro, J. Baier, C. Ruz, A. Soto},\n title\t\t= {How a General-Purpose Commonsense Ontology can Improve\n\t\t Performance of Learning-Based Image Retrieval},\n booktitle\t= {{IJCAI}},\n year\t\t= {2017},\n abstract\t= {The knowledge representation community has built\n\t\t general-purpose ontologies which contain large amounts of\n\t\t commonsense knowledge over relevant aspects of the world,\n\t\t including useful visual information, e.g.: \"a ball is used\n\t\t by a football player\", \"a tennis player is located at a\n\t\t tennis court\". Current state-of-the-art approaches for\n\t\t visual recognition do not exploit these rule-based\n\t\t knowledge sources. Instead, they learn recognition models\n\t\t directly from training examples. In this paper, we study\n\t\t how general-purpose ontologies---specifically, MIT's\n\t\t ConceptNet ontology---can improve the performance of\n\t\t state-of-the-art vision systems. As a testbed, we tackle\n\t\t the problem of sentence-based image retrieval. Our\n\t\t retrieval approach incorporates knowledge from ConceptNet\n\t\t on top of a large pool of object detectors derived from a\n\t\t deep learning technique. In our experiments, we show that\n\t\t ConceptNet can improve performance on a common benchmark\n\t\t dataset. Key to our performance is the use of the ESPGAME\n\t\t dataset to select visually relevant relations from\n\t\t ConceptNet. Consequently, a main conclusion of this work is\n\t\t that general-purpose commonsense ontologies improve\n\t\t performance on visual reasoning tasks when properly\n\t\t filtered to select meaningful visual relations. },\n url\t\t= {https://arxiv.org/abs/1705.08844}\n}\n\n","author_short":["R. Toro, J. B.","C. Ruz, A. S."],"key":"toro:etal:2017","id":"toro:etal:2017","bibbaseid":"rtoro-cruz-howageneralpurposecommonsenseontologycanimproveperformanceoflearningbasedimageretrieval-2017","role":"author","urls":{"Paper":"https://arxiv.org/abs/1705.08844"},"metadata":{"authorlinks":{}}},"bibtype":"inproceedings","biburl":"https://raw.githubusercontent.com/ialab-puc/ialab.ing.puc.cl/master/pubs.bib","dataSources":["sg6yZ29Z2xB5xP79R"],"keywords":[],"search_terms":["general","purpose","commonsense","ontology","improve","performance","learning","based","image","retrieval","r. toro","c. ruz"],"title":"How a General-Purpose Commonsense Ontology can Improve Performance of Learning-Based Image Retrieval","year":2017}