Semantic Intersections between Artificial and Artistic Vision

Pilar Rosado, Eva Figueras y Ferran Reverter

BRAC - Barcelona Research Art Creation. Vol. 2 No. 1, February 2014, pp. 1-54



In our project we have approached the difficulties of automatic classification of images on which the conception and design of sculptor M. Planas artistic production are based. This artist constantly generates images in his creative process. The methodology used is based on local characteristics. In order to build up a visual vocabulary for basing image description on, we followed a procedure similar to the one used in automatic text analysis. The method is known as the “Bag-of-Words” (BOW) model because every document is represented as a distribution of frequencies of the words in the text, without considering the syntactic relationships among them. In the sphere of images we refer to “Bag-of-Visual Terms” (BOV) representations. This approach consists in analysing images as a group of regions, describing only their appearance without taking into account their spatial structure. To overcome the disadvantages of polysemy and synonymy that this methodology has associated, we have used probabilistic latent semantic analysis (PLSA), that detects underlying topic in images. The outcomes are promising, the described cataloguing method may provide the artist with new viewpoints for future works.