[CESERH] CI La généalogie des formes dans la perspective du Deep Learning 7/11/2019 [2/11]

[CESERH] CI La généalogie des formes dans la perspective du Deep Learning 7/11/2019 [2/11]

[CESERH] CI The genealogy of forms from the Deep Learning perspective 7/11/2019 [2/11]

by Maria Giulia Dondero · View the channel on PSI

Formal & Physical Sciences Computing & Cybersecurity UXApplied computingUXADigital Humanities
🎙 Maria Giulia Dondero 👥 178 📅 November 18, 2019 ⏱ 39 min 👁 62 📄 expert opinion 🧭 2026-10-10
Available in: English (current) Français

Keywords

genealogy of formsWarburgplastic featuresneural networksimage segmentation

Summary

This lecture by Maria Giulia Dondero, presented at the Centre de Sémiotique & Rhétorique of ULiège, explores the intersection of semiotics and deep learning for the analysis of visual forms. The speaker introduces two research projects: one on semiotic indexing and another on computational image analysis. She discusses the genealogy of forms, referencing Aby Warburg’s Mnemosyne Atlas and the concept of Pathosformel, and proposes that deep learning can assist in detecting formal and motivic similarities across large image corpora. The talk addresses methodological challenges, such as the distinction between classification based on metadata versus visual features, and the importance of technique in art analysis. She also critiques the limitations of deep learning in understanding abstract art. Throughout, Dondero emphasizes the need for a semiotic framework to interpret the outputs of neural networks, bridging the gap between computational pattern recognition and humanistic interpretation.

142 words

Critical Evaluation

Value of the Information & Strength of the Argument

The value of the information lies in its interdisciplinary approach, bringing together semiotic theory and computational methods. Dondero argues that deep learning can reveal formal genealogies that influence visual culture, but stresses the necessity of a semiotic analysis to understand these patterns meaningfully. The argumentation is sophisticated, drawing on historical art theory (Warburg) and contemporary computational techniques. However, the lecture is dense and assumes prior knowledge; the chain of reasoning is not always explicit, and the poor transcript hampers clarity.

Scientific Rigor, Source Quality, Title Accuracy

Dondero references key figures in art history and semiotics, such as Aby Warburg, Georges Didi-Huberman, and René Thom, indicating a solid scholarly basis. The sources mentioned are academic but no specific citations or URLs are provided in the description. The title accurately matches the content, and the lecture maintains a focused scope. The lack of verifiable external sources and the technical depth may limit accessibility, but the internal coherence is strong.

166 words

Title / Content Match

The title accurately reflects the content: a focus on the genealogy of forms through the lens of deep learning, with explicit connections to Warburg's work and computational image analysis.

Quality & Reliability

7/10

The talk is presented by a recognized researcher and covers a niche but coherent intersection of semiotics and deep learning. The transcript is poor quality, which limits detailed verification, but the intellectual framework and references align with established academic discourse.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

  • Aby Warburg, Mnemosyne Atlas — Referenced as a foundational visual archive for studying genealogies of forms.
  • Georges Didi-Huberman, on pathosformel — Mentioned in relation to the survival of motifs and emotional formulas.
  • René Thom, semiotics and morphology — Cited for his theories on form and dynamics.
  • Marco Cardinali, digital tools in art history — Mentioned in context of technical analysis of paintings.

Contribution & Novelties

The talk brings a new perspective by applying deep learning to the genealogical analysis of visual forms, bridging Warburg’s iconology with contemporary AI. It highlights the importance of considering the material and technical aspects of art creation, and proposes that computational methods can enhance but not replace semiotic interpretation.

Pour aller plus loin :

  • Aby Warburg — Overview of Warburg’s life and the Mnemosyne Atlas.
  • Deep learning — General introduction to deep learning techniques.
  • Computer vision — Covers the field of image analysis with computers.

85 words

Radar Profile

The radar profile shows high technical level and moderate information quantity, with lower reliability scores due to the poor transcript and lack of verifiable sources. The balance suggests a specialized, intellectually stimulating but somewhat inaccessible presentation.

Reliability 6/10