WICED 2017

Permanent URI for this collection

https://diglib.eg.org/handle/10.2312/2631243

Browse

Browsing WICED 2017 by Subject "I.2.10 [Artificial Intelligence]"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Inferring the Structure of Action Movies
    (The Eurographics Association, 2017) Potapov, Danila; Douze, Matthijs; Revaud, Jérôme; Harchaoui, Zaid; Schmid, Cordelia; William Bares and Vineet Gandhi and Quentin Galvane and Remi Ronfard
    While important advances were recently made towards temporally localizing and recognizing specific human actions or activities in videos, efficient detection and classification of long video chunks belonging to semantically-defined categories remains challenging. Examples of such categories can be found in action movies, whose storylines often follow a standardized structure corresponding to a sequence of typical segments such as ''pursuit'', ''romance'', etc. We introduce a new dataset, Action Movie Franchises, consisting of a collection of Hollywood action movie franchises. We define 11 non-exclusive semantic categories that are broad enough to cover most of the movie footage. The corresponding events are annotated as groups of video shots, possibly overlapping. We propose an approach for localizing events based on classifying shots into categories and learning the temporal constraints between shots. We show that temporal constraints significantly improve the classification performance. We set up an evaluation protocol for event localization as well as for shot classification, depending on whether movies from the same franchise are present or not in the training data.
  • Thumbnail Image
    Item
    Using ECPs for Interactive Applications in Virtual Cinematography
    (The Eurographics Association, 2017) Wu, Hui-Yin; Li, Tsai-Yen; Christie, Marc; William Bares and Vineet Gandhi and Quentin Galvane and Remi Ronfard
    This paper introduces an interactive application of our previous work on the Patterns language as creative assistant for editing cameras in 3D virtual environments. Patterns is a set of vocabulary, which was inspired by professional film practice and textbook terminology. The vocabulary allows one to define recurrent stylistic constraints on a sequence of shots, which we term ''embedded constraint pattern'' (ECP). In our previous work, we proposed a solver that allows us to search for occurrences of ECPs in annotated data, and showed its use in automated analysis of story and emotional elements of film. This work implements a new solver that interactively propose framing compositions from an annotated database of framings that conform to the user-applied ECPs. We envision this work to be incorporated into tools and interfaces for 3D environments in the context of film pre-visualisation, film or digital arts education, video games, and other related applications in film and multimedia.