EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Foveated Retinotopy Improves Classification and Localization in Convolutional Neural Networks

Jean-Nicolas Jérémie iD, Emmanuel Daucé iD, Laurent U Perrinet iD

Vision · 30.03.2026 · 10 (2) · 17

DOI: 10.3390/vision10020017 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Vision, online erschienen am 30.03.2026. Verfasst von Jean-Nicolas Jérémie und 2 weiteren Autor:innen. Beteiligte Einrichtung: Institut de Neurosciences de la Timone, Aix-Marseille Université, CNRS UMR 7289, 13005 Marseille, France. Die Arbeit nennt 63 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Agence Nationale de la Recherche, ANR https://ror.org/00rbzpz17, French government.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

From falcons spotting prey to humans recognizing faces, the ability to rapidly process visual information depends on a foveated retinal organization that provides high-acuity central vision while preserving low-resolution peripheral vision. This organization is conserved along early visual pathways, yet remains under-explored in machine learning. Here, we examine the impact of embedding a foveated retinotopic transformation as a preprocessing layer on convolutional neural networks (CNNs) for image classification. By applying a log-polar mapping to off-the-shelf models and retraining them, we achieve comparable accuracy while improving robustness to scale and rotation. We demonstrate that this architecture is highly sensitive to shifts in the fixation point and that this sensitivity provides an effective proxy for defining saliency maps that facilitate object localization. Our results demonstrate that foveated retinotopy encodes prior geometric knowledge, providing a solution for visual searches and a meaningful classification robustness and localization trade-off. These findings provides a proof of concept in order to connect principles of biological vision with artificial networks, suggesting new, robust and efficient approaches for computer vision systems.

Schlagwörter

Stichwörter der Autor:innen

NeuroAIconvolutional neural networksfoveated visionneuromorphic transformationtransfer learningvisual categorisation

Autor:innen und Institutionen

  • Institut de Neurosciences de la Timone, Aix-Marseille Université, CNRS UMR 7289, 13005 Marseille, France.
  • Ecole Centrale Méditerranée, Technopôle Château-Gombert, 13013 Marseille, France.

ORCID-Profile: Jean-Nicolas Jérémie, Emmanuel Daucé, Laurent U Perrinet

Förderung

  • Agence Nationale de la Recherche
  • ANR https://ror.org/00rbzpz17
  • French government

Zitieren

APA

Jérémie, J. N., Daucé, E., & Perrinet, L. U. (2026). Foveated Retinotopy Improves Classification and Localization in Convolutional Neural Networks. Vision, 10(2), 17. https://doi.org/10.3390/vision10020017

Vancouver

Jérémie JN, Daucé E, Perrinet LU. Foveated Retinotopy Improves Classification and Localization in Convolutional Neural Networks. Vision. 2026;10(2):17. doi:10.3390/vision10020017

BibTeX
@article{Jrmie2026,
  author = {Jérémie, Jean-Nicolas and Daucé, Emmanuel and Perrinet, Laurent U},
  title = {Foveated Retinotopy Improves Classification and Localization in Convolutional Neural Networks},
  journal = {Vision},
  year = {2026},
  volume = {10},
  number = {2},
  pages = {17},
  doi = {10.3390/vision10020017}
}
Kostenlos starten →