Erika Scaletta

  • PhD researcher in Computer Science
  • Founder of Sei Arte
  • Department of Mathematics and Computer Science, University of Catania
  • Visiting PhD researcher at IDLab, Ghent University
  • Supervisor: Prof. Emiliano Alessio Tramontana

I teach machines to read places from above: what grows in a field, how a city is built, and where a road becomes dangerous for cyclists.

About

I am a researcher in Computer Science who came to code through literature. After a B.A. in Modern Literature and an M.Sc. in Digital Humanities at the University of Catania, where I built a digital library of Sicilian and co-developed DanteGPT, I started a PhD in Computer Science in December 2024, supervised by Prof. Emiliano Alessio Tramontana and co-funded by the National Centre for HPC, Big Data and Quantum Computing. Reading texts closely taught me to read places the same way. Today I train vision models on aerial images of Sicily, turn Christopher Alexander's urban patterns into rules computed on OpenStreetMap data and, as a visiting PhD researcher at IDLab in Ghent, look for the road features that make cycling races dangerous.

Research interests

  • computer vision for aerial and satellite imagery
  • object detection and segmentation
  • urban computing
  • Christopher Alexander's design patterns
  • geospatial data
  • sport data science
  • cycling safety

Outside the lab: karate, black belt 3rd dan, athlete and instructor for children.

Research

Three lines, in order of time. Move the lens over each image to see what the model sees.

Since 2025

Aerial and satellite imagery

Computer vision and deep learning, object detection and segmentation, on real aerial and satellite images of Sicily, on a dataset I built: by May 2026, 605 images, 3,926 annotated instances and 11 categories. A hybrid YOLO-C3 pipeline, a pixel-based classifier plus YOLO, designed for drones. And a study of what LLMs can and cannot see in aerial images.

YOLO-C3 output on an aerial image.

Three images from the dataset, Sicily. Image source: Google Maps.

Since 2026

Urban pattern analysis

Urban computing software that translates Christopher Alexander's design patterns into computable rules on open geospatial data, for the automatic evaluation of urban areas. Pattern 11, Local Transport Areas, became seven measurable rules, applied to 23 cities in City Explorer. In progress: the shape of junctions, T versus X, read from aerial images for Pattern 50, T Junctions.

Caltagirone, two candidate zones. Map data © OpenStreetMap contributors.

Ghent, 2026

Sport data science: cycling safety

As a visiting PhD researcher at IDLab, Ghent University, with Prof. Steven Verstockt: the safety of cycling races, in the line the group develops in collaboration with the Union Cycliste Internationale (UCI). AI models that automatically find the dangerous points of race routes.

Koppenberg, Flanders. Route from OpenStreetMap; the markers are illustrative, not IDLab results. Map data © OpenStreetMap contributors.

Publications

Most recent first.

    Experience

    Contact

    Email

    erika.scaletta@phd.unict.it

    Profiles

    Funding

    PhD scholarship co-funded by the National Centre for HPC, Big Data and Quantum Computing (ICSC), Spoke 2, PNRR. Funded by the European Union, NextGenerationEU.