Cooking action recognition with iVAT: An interactive video annotation tool

Abstract

Within a video recipe, we are interested in locating and annotating the various ingredients, kitchenwares and relevant cooking actions. To this end we have developed the iVAT interactive video annotation tool to support manual, semi-automatic and automatic annotations obtained on the basis of the interaction of the user with various detection algorithms. The tool integrates versions of computer vision algorithms, specifically adapted to work in an interactive and incremental learning framework. iVAT has been developed to annotate video recipes, but it can be easily adapted and used to annotate videos from different domains as well. In this paper we present some results with respect to the task of cooking action recognition. © 2013 Springer-Verlag.