Visual Mining

Visual Mining

Discovering visual patterns and typologies through categorization and semiotic analysis of images and videos (memes, photos, illustrations, etc.) using machine learning.

Identifying viral images and reproductions (versioning) allows for a semiotic and statistical understanding of the most impactful visual content. This approach highlights the dominant aesthetics and imaginaries influencing contemporary visual culture. Concurrently, detecting emerging imaginaries and alternative aesthetic trends offers the opportunity to anticipate cultural shifts and adapt visual communication strategies accordingly.

Usage and applications

  • DECODE
    Unraveling representations and cultural codes embedded in visual content
  • REVEAL
    Discerning the value of locations and imaginaries, identifying gaps or overlooked areas
  • DECIDE
    Investing in lesser-known aesthetics that social networks may bring to light, defining the aesthetic and cultural codes to mobilize

Methodology

  • Statistical Sorting: Leveraging supervised semantic analysis for data organization
  • Automated Image Annotation: Employing CLIP for efficient image tagging and categorization
  • Semantic Mapping: Creating a visual representation map based on the BERT model (under development), offering an insightful understanding of visual content and its underlying themes