In order to semantically label visual objects in a large amount of images, we propose a new approach which is fast and accurate. The developed automatic object labeling technique relies on the embedded double matching of local visual descriptors detected in the example and candidate images to semantically tag the candidate image with the label associated to the example image, leading thus to the automatic annotation of the candidate image. Our system has been successfully tested on different standard datasets and is compatible with online applications.
|Number of pages||2|
|Publication status||Published - 13 Dec 2012|
|Event||Vision and Language Net Workshop on Vision and Language 2012 - The Edge Conference Centre, University of Sheffield, Sheffield, United Kingdom|
Duration: 13 Dec 2012 → 14 Dec 2012
|Conference||Vision and Language Net Workshop on Vision and Language 2012|
|Period||13/12/12 → 14/12/12|