Friday, 25 de April de 2014

Ficha del recurso:

Fuente:

Vínculo original

Lugar:

Capri ,  Italy

Inicio:

Monday 21 de May de 2012

Fin:

Monday 21 de May de 2012

Coste:

Desconocido

Última actualización:

Tuesday, 17 de April de 2012

Entrada en el observatorio:

Tuesday, 17 de April de 2012

Idioma:

Inglés

Archivado en:

First International Workshop on Visual Interfaces for Ground Truth Collection in Computer Vision Applications

The importance of having image/video database containing high quality ground truth annotations generated by humans for a variety of computer vision applications is recognized by the whole machine vision community. Indeed, one of the most significant efforts during the evaluation process is represented by the devel- opment of accurate truth and comparing this truth to the decision of image and video processing applications. For example, datasets with ground truth labels are necessary for supervised learning of object categories. However, the cost of providing labelled data, which implies asking a human to examine images and provide labels, becomes impractical as training sets grow.

The computer vision community has mainly directed its attention to develop methods for collecting large scale datasets by exploiting the collaborative ef- fort of a large population of annotators. Although there exist functional, task- oriented requirements for tools supporting ground truth labeling, the research still lacks in following human computer interaction paradigm to develop user- oriented tools. In fact, annotators, must be at the centre of such tools since the ground truth needs to be established by humans who, for example, on a video sequence of five minutes captured at 30 fps, have to annotate manually a total of 9000 frames. Therefore, tools for ground truth annotation must be user-oriented, providing advanced visual interfaces able to speed-up the process of ground truth creation by helping users, through the integration of computer vision methods and multi-users ground truth annotations, to finish the task in a reasonable amount of time with less effort.

The workshop aims at bringing together practitioners and researchers both in computer vision and in HCI and in multimedia research to share ideas and experiences in designing and implementing visual interfaces for ground truth data generation, thus promoting interdisciplinary research between computer vision, HCI and multimedia research.

The workshop will be held in conjunction with the 11th International Working Conference on Advanced Visual Interfaces (AVI2012) and is supported by the European Commission funded project “Fish4Knowedge” Seventh Framework Programme [FP7/2007-2013] grant agreement 257024.