CVIU: Large Scale Data-Driven Evaluation in Computer Vision Call for Papers

Call for Papers Special Issue on “Large Scale Data-Driven Evaluation in Computer Vision” Computer Vision and Image Understanding Journal (Elsevier) http://www.journals.elsevier.com/computer-vision-and-image-understanding/ Deadline for submissions: 30th November 2013 =============== Description =============== In the development of computer vision and image analysis methods, a fundamental role is played by the availability of large scale visual datasets providing a wide coverage of different object classes, scenarios and environments. These are used for: 1) training machine-learning algorithms, which have been largely adopted for computer vision, but still strongly suffer the lack of comprehensive, large-scale training data, and 2) evaluating the performance of algorithms, which has to provide enough evidence, to the developers that a method works well in the targeted conditions. Furthermore, large scale datasets would also allow the development of “visual-data driven” approaches, which in the long run could be independent from human annotations. However, the main limitation to collect such datasets is the daunting amount of time and human effort necessary to annotate accurately them or part of them; in fact, it has been estimated that labeling an image may take from two to thirty minutes, depending on the task, and this is, obviously, even worse in case of videos. In addition, issues such as annotation quality, confidence level, missing data, etc. arise, especially, when dealing with small sets of visual contents as most available datasets which are produced as the result of efforts of single research groups who have manually annotated (or have paid to get the annotations using platforms such as Amazon Mechanical Turk) them with a specific task in mind. On the other hand, the existing large scale datasets suffer from many limitations, for example, absent, incomplete or low-quality annotations to lack of broad coverage to interoperability issues. An alternative experienced way to generate a large set of annotated data has been to create synthetic datasets, however, these may not able to reproduce the real-world conditions and variations, thus simplifying the problem and making the synthetically rendered sequences unfit to evaluate real conditions. This special issue aims, therefore, at presenting and reporting the most recent efforts: 1) to support automatic or semi-automatic generation of large scale datasets together with annotations, 2) to integrate existing datasets by investigating harvesting and representation schema matching approaches, 3) to exploit big visual data and the Internet crowd to overcome the lack of annotated datasets and 4) to develop “data-driven” approaches also able to evaluate algorithms’ performance with limited or no ground truth data. ================== Topics of Interest ================== Submissions are encouraged, but not limited, to the following topics: - Computer vision and machine learning methods for supporting humans in the task of annotating visual datasets more efficiently. - Computer vision and machine learning methods to combine annotations in the form of both textual labels and graphical items. - Crowdsourcing methods to generate fast and high quality annotations. - Methods for performance evaluation with limited or no ground truth data - Big visual data for supporting unsupervised (and its variants) learning of visual tasks. - Interactive and collaborative annotation tools. - Framework for sharing datasets, ground truths, features, algorithms and tools. - Ontologies and web semantic approaches for dataset modeling, sharing and harvesting. ========================= Submission Instructions ========================= Submissions to the special issue must include new, unpublished, original research. Papers must be original and have not been published or submitted elsewhere. All papers must be written in English. Prospective authors should submit an electronic copy of their complete manuscript through the journal submission system at http://ees.elsevier.com/cviu/ according to the submission schedule. After login the authors should choose the Special Issue “Large Scale Data-Driven Evaluation in Computer Vision” in the "Choose Article Type" section. All submissions will undergo initial screening by the Editors for fit to the theme of the Special Issue and prospects for successfully negotiating the review process. =============== Important Dates =============== Submission of papers: 30 November 2013 Acceptance/Revision notification: 28 February 2014 Revised manuscript due: 15 April 2014 Final acceptance notification: 30 June 2014 ========================== Special Issue Guest Editors ========================== Dr. Concetto Spampinato, University of Catania, Italy -- cspampin@dieei.unict.it Dr. Bas Boom, University of Edinburgh, UK -- bboom@inf.ed.ac.uk Prof. Benoit Huet, EURECOM, France -- Benoit.Huet@eurecom.fr