Yannis Skarpelos — Augmented Images (ULiège)

Yannis Skarpelos — Augmented Images (ULiège)

Title history
  • Yannis Skarpelos — Augmented Images (ULiège) — observed on 20261026-1637 UTC (YouTube)
🎙 Yannis Skarpelos 👥 180 📅 October 15, 2022 ⏱ 32 min 👁 120 📄 original study 🧭 2026-10-26
Available in: English (current) Français

Keywords

plastic signsentropycolor palettelarge visual corporasociology

Summary

Yannis Skarpelos presents a computational semiotic analysis of large visual corpora, focusing on Greek music album covers (1960-2016) and a comparison with a women’s magazine (Romanzo). He argues that while conventional semiotics excels at interpreting a few iconic images, it fails to handle thousands of objects. Using machine learning (convolutional neural networks, t-SNE) and computer vision, he analyzes two plastic signs: entropy and color. Entropy measures visual order vs. fragmentation, which he links to societal value shifts, e.g., increasing entropy aligns with post-materialist values per Inglehart. Color analysis uses K-means clustering to extract a palette of 12 centroids, based on Berlin and Kay’s basic color terms. He identifies a three-tier structure of color stability: constant colors (white, black, orange, yellow, blue), transitional colors (beige, brown, light blue, green, greys) that changed notably from the 80s to 90s, and transient colors linked to fads. He correlates these shifts with Greek political history—military dictatorship, EU accession, economic crisis. He also examines magazine covers, showing similar color trends, e.g., green declining after the 80s. The talk demonstrates how computational analysis can relate low-level visual features to intangible social realities like values and historical turning points.

192 words

Critical Evaluation

Value of the Information & Strength of the Argument

The value of this information lies in its innovative combination of semiotic theory with large-scale quantitative analysis, addressing a gap in visual studies that typically focuses on exceptional cases. The argumentation is solid: the speaker establishes a theoretical basis (Peirce, Barthes, Groupe μ) and then presents a concrete methodological pipeline (feature extraction, clustering) with clear results. He is cautious about interpretation, acknowledging that correlations with societal changes are tentative and require further fine-graining. The comparative analysis across different media (album covers vs. magazine covers) strengthens the claim that plastic signs can reflect broader cultural patterns. The use of entropy as a measure is justified, and the color analysis is grounded in established color terminology. Overall, the argument is coherent and well-supported, though the specificity of the Greek context limits generalizability without replication.

Scientific Rigor, Source Quality, Title Accuracy

The talk demonstrates scientific rigor by grounding the analysis in established semiotic and sociological theories, citing Peirce, Barthes, Inglehart, Manovich, Berlin and Kay, and Groupe μ. The methodology is described in sufficient detail (corpus size, clustering parameters, color distance formulas) to be reproducible. The title ‘Augmented Images’ refers to the symposium theme, while the actual talk title (as implied in description) is ‘Modeling Plastic Signs in Big Visual Corpora’—the video title does not reflect this specific topic, but it accurately identifies the speaker and event. No external sources are hyperlinked in the description, so the only sources are those named in the talk. The absence of links limits immediate verification, but the academic context (conference) and the speaker’s credentials lend credibility.

268 words

Title / Content Match

The video title indicates the speaker and the symposium name ('Augmented Images') but does not mention the specific topic of the talk. The title is thus only partially indicative of the content, but it correctly situates the speaker and event.

Quality & Reliability

8/10

Presentation of original research with a clear methodological framework, combining semiotics and computational analysis. The speaker works with a near-complete corpus of Greek album covers and provides quantitative results (entropy, color palettes) that are interpretable through sociological theory. Limitations include the lack of external verification of the specific data and the reliance on self-reported methodology, but the approach appears consistent and credible.

Key Moments

Cited Sources

  • Peirce, C.S. – Semiotics — Reference to Peirce's concepts of representation, representamen, and object, as well as qualia as firstness.
  • Barthes, R. – Connotation — Mention of Barthes' perfect analogon and connotation as second-order signification.
  • Inglehart, R. – Post-materialism — Used to explain the shift from materialist to post-materialist values and its correlation with entropy changes.
  • Manovich, L. – Democratic art history — Insisted on studying all production rather than just iconic outliers, influencing the corpus approach.
  • Berlin & Kay – Basic Color Terms — Basis for selecting 12 color centroids, referring to the 11 basic color terms in Western languages.
  • Groupe μ – Plastic signs — Mention of plastic signs as low-level signifiers (color, texture) and semi-symbolic systems.

Concurring Sources

  • Inglehart's post-materialism studies — Finds a global shift toward self-expression values, consistent with observed entropy increase.
  • Manovich's 'democratic art history' — Advocates analyzing total production, aligning with the speaker's corpus approach.

Dissenting Sources

  • Group μ's critique of entropy use — The speaker notes that the teleological use of entropy by Arnheim (and criticized by Groupe μ) differs from the statistical entropy used in software, requiring careful interpretation.

Contribution & Novelties

The originality lies in applying computational methods to a near-complete national corpus (44,000+ album covers) and linking plastic signs (entropy, color) to sociological changes over 50 years. This moves beyond point analyses of iconic images to a macro-level quantitative semiotics. The three-tier structure of color stability offers a novel way to interpret cultural constants and shifts. The comparison with magazine covers extends the method across media, suggesting broader applicability. The talk also contributes to discussions on how big visual data can be analyzed semiotically without losing theoretical depth.

Pour aller plus loin :

145 words

Radar Profile

The radar profile shows high scores in quantity and technical level, indicating a data-rich and methodologically advanced presentation. Quality and reliability are slightly lower, reflecting the inherent uncertainty in interpreting correlations. The overall shape suggests a technical talk with strong empirical grounding, though theoretical interpretation carries some speculative elements.

Reliability 8/10