< Back to list

Towards an Inpainting Framework for Visual Cultural Heritage

conferencePaper

DOI:10.1109/JEEIT.2019.8717470
Authors: Jboor Nesreen Hamdallah / Belhi Abdelhak / Al-Ali Abdulaziz Khalid / Bouras Abdelaziz / Jaoua Ali

Extracted Abstract:

— Cultural heritage takes an important part in defining the identity and the history of a civilization or a na- tion. Valuing and preserving this heritage is thus a top prior- ity for governments and heritage institutions. Through this paper, we present an image completion (inpainting) approach adapted for the curation and the completion of damaged art- work. Our approach uses a set of machine learning techniques such as Generative Adversarial Networks which are among the most powerful generative models that can be trained to generate realistic data samples. As we are focusing mostly on visual cultural heritage, the pipeline of our framework has many optimizations such as the use of clustering to optimize the training of the generative part to ensure a better perfor- mance across a variety of cultural data categories. The exper- imental results of our framework are promising and were val- idated on a dataset of paintings. Keywords— Image Inpainting, Generative Adversarial Net- works, Deep Learning, Cultural Heritage I.

Level 1: Include/Exclude

  • Papers must discuss situated information visualization* (by Willet et al.) in the application domain of CH.
    *A situated data representation is a data representation whose physical presentation is located close to the data’s physical referent(s).
    *A situated visualization is a situated data representation for which the presentation is purely visual – and is typically displayed on a screen.
  • Representation must include abstract data (e.g., metadata).
  • Papers focused solely on digital reconstruction without information visualization aspects are excluded.
  • Posters and workshop papers are excluded to focus on mature research contributions.
Show all meta-data