Publications
Here is a selection of my publications, organized by research area:
For a complete list, please visit my Google Scholar profile.
đ§ General
PLENARY: Explaining black-box models in natural language through fuzzy linguistic summaries
Authors: Katarzyna Kaczmarek-Majer, Gabriella Casalino, Giovanna Castellano, Monika Dominiak, Olgierd Hryniewicz, Olga KamiĆska, Gennaro Vessio, Natalia DĂaz-RodrĂguez
Published in: Inf Sci, 2022Summary: PLENARY is an explainable AI approach that uses fuzzy linguistic summaries to translate model explanations into natural language, making predictive model outputs more understandable.
MLOps: A Taxonomy and a Methodology
Authors: Matteo Testi, Matteo Ballabio, Emanuele Frontoni, Giulio Iannello, Sara Moccia, Paolo Soda, Gennaro Vessio
Published in: IEEE Access, 2022Summary: This paper reviews MLOps literature, proposes a taxonomy, and introduces a ten-step pipeline to streamline ML deployment in industry, aiming to standardize practices for effective ML adoption.
đš Digital Humanities
A Deep Learning Approach to Clustering Visual Arts
Authors: Giovanna Castellano, Gennaro Vessio
Published in: IJCV, 2022Summary: We propose a deep learning-based approach for clustering artworks. It uses a pre-trained convolutional network to extract features and a deep embedded clustering model to map these features into clusters. This method effectively identifies patterns in visual art, aiding tasks like visual link retrieval and historical knowledge discovery in painting datasets.
Leveraging Knowledge Graphs and Deep Learning for automatic art analysis
Authors: Giovanna Castellano, Vincenzo Digeno, Giovanni Sansaro, Gennaro Vessio
Published in: KBS, 2022Summary: We present ArtGraph, a knowledge graph that enriches fine art analysis by linking data from WikiArt and DBpedia. It supports knowledge discovery and enables a new method for fine art classification, where ArtGraph embeddings enhance deep learning models, bridging Humanities and Computer Science.
đ Drone Vision
Weed mapping in multispectral drone imagery using lightweight vision transformers
Authors: Giovanna Castellano, Pasquale De Marinis, Gennaro Vessio
Published in: Neurocomputing, 2023Summary: This paper introduces a lightweight Vision Transformer for weed mapping in agriculture, using multispectral drone images to accurately segment crops and weeds, enhancing treatment prioritization and crop yield.
Density-based clustering with fully-convolutional networks for crowd flow detection from drones
Authors: Giovanna Castellano, Eugenio Cotardo, Corrado Mencar, Gennaro Vessio
Published in: Neurocomputing, 2023Summary: This paper proposes a drone-based crowd flow detection method using a fully-convolutional network to cluster and track crowd movements in video sequences. Tested on the VisDrone datasets, this approach effectively analyzes crowd behavior, opening new possibilities for high-level crowd monitoring from drones.
đ„ e-Health
Automated detection of Alzheimerâs disease: a multi-modal approach with 3D MRI and amyloid PET
Authors: Giovanna Castellano, Andrea Esposito, Eufemia Lella, Graziano Montanaro, Gennaro Vessio
Published in: Sci Rep, 2024Summary: We present a multi-modal deep learning approach for Alzheimerâs diagnosis, combining 3D MRI and amyloid PET scans to outperform single-modality models. Achieving top results on the OASIS-3 cohort, the model also highlights key brain regions linked to AD, aiding disease understanding.
Forecasting and what-if analysis of new positive COVID-19 cases during the first three waves in Italy
Authors: Serena De Ruvo, Gianvito Pio, Gennaro Vessio, Vincenzo Volpe
Published in: MBEC, 2023Summary: This study combines epidemiological, mobility, and restriction data with machine learning to predict COVID-19 cases in Italy during the first three pandemic waves. The model accurately forecasts new cases, achieving a WAPE of 5.75% nationally, and supports what-if analyses, suggesting targeted restrictions are more effective than total lockdowns.
đ Miscellaneous
Pathways to success: a machine learning approach to predicting investor dynamics in equity and lending crowdfunding campaigns
Authors: Rosa Porro, Thomas Ercole, Giuseppe PipitĂČ, Gennaro Vessio, Corrado Loglisci
Published in: J Intell Inf Syst, 2024Summary: This paper analyzes investor behavior in Italian crowdfunding, using ML models to predict campaign success and introducing new datasets and metrics. The insights aim to enhance campaign strategies and platform decision-making.
ROULETTE: A neural attention multi-output model for explainable Network Intrusion Detection
Authors: Giuseppina Andresini, Annalisa Appice, Francesco Paolo Caforio, Donato Malerba, Gennaro Vessio
Published in: ESWA, 2022Summary: ROULETTE is an explainable neural model for network intrusion detection, using attention and multi-output learning to improve accuracy and interpretability. Tested on NSL-KDD and UNSW-NB15 datasets, it effectively classifies network traffic data.