Portrait of Gabriele RosatiPortrait of Gabriele Rosati

Gabriele Rosati

Research Fellow in computer vision and deep learning for medical and industrial image analysis at UNIMORE.

  • Modena, Italy
  • M.Sc. in AI Engineering
  • Member of AImageLab, Department of Engineering "Enzo Ferrari"

Links

About

I am a Research Fellow at the University of Modena and Reggio Emilia and a member of AImageLab, in the Department of Engineering “Enzo Ferrari”.

My work connects computer vision, deep learning, and applied image analysis across medical and industrial settings. I have worked on renal whole-slide images, glomeruli detection and classification, CBCT segmentation benchmarks for dental and maxillofacial imaging, and visual inspection systems for pharmaceutical vial quality control.

I am especially interested in practical AI systems: data pipelines, robust evaluation protocols, medical imaging workflows, weakly supervised learning, and models that can move from research code toward useful tools.

Technical Skills

  • Python
  • PyTorch
  • Deep learning
  • Computer vision
  • Medical imaging
  • Whole-slide images
  • CBCT volumes
  • YOLO
  • ResNet
  • Multi-instance learning
  • Image classification
  • Segmentation
  • Data curation
  • Evaluation protocols
  • Git and GitHub
  • Django

Selected Publications

Multi-Structure Segmentation in CBCT Volumes: the ToothFairy2 ChallengeMulti-Structure Segmentation in CBCT Volumes: the ToothFairy2 Challenge

Multi-Structure Segmentation in CBCT Volumes: the ToothFairy2 Challenge

6/1/2026
Publication
Medical Image Analysis 2026 article presenting the ToothFairy2 challenge, a public benchmark for 42-class multi-structure segmentation in maxillofacial CBCT volumes.
ToothFairy3: Scaling CBCT Maxillofacial Segmentation to 77 Classes with U-Mamba2ToothFairy3: Scaling CBCT Maxillofacial Segmentation to 77 Classes with U-Mamba2

ToothFairy3: Scaling CBCT Maxillofacial Segmentation to 77 Classes with U-Mamba2

6/1/2026
Publication
MICCAI 2026 work extending the ToothFairy CBCT benchmark to 77 maxillofacial classes, adding fine pulp and neurovascular structures and introducing U-Mamba2 for efficient 3D segmentation.
Identifying Impurities in Liquids of Pharmaceutical VialsIdentifying Impurities in Liquids of Pharmaceutical Vials

Identifying Impurities in Liquids of Pharmaceutical Vials

12/1/2024
Publication
First-author ICPR 2024 paper by Gabriele Rosati on deep-learning inspection of pharmaceutical liquid vials, with ResNet-18 and multi-instance learning trained on rotating-vial image sequences.

View all research and projects

Software and Thesis

ESKD Prediction Web ApplicationESKD Prediction Web Application

ESKD Prediction Web Application

1/1/2026
Software
Full-stack clinical AI web application for end-stage kidney disease risk prediction in IgA nephropathy, combining model development, survival analysis, explainability, Django deployment, and medical decision-support UI.

Glomerular Feature Classification from WSI

7/16/2025
Software
ResNet-18 pipeline for classifying glomerular location, appearance, and distribution features from extracted renal WSI glomeruli, with WSI-level aggregation rules.
Glomeruli Detection in Whole-Slide ImagesGlomeruli Detection in Whole-Slide Images

Glomeruli Detection in Whole-Slide Images

11/19/2024
Software
YOLOv8-based renal pathology pipeline for detecting glomeruli in whole-slide images, including tissue preprocessing, patch extraction, annotation handling, training, and reconstruction to slide coordinates.

Prediction of Kidney Failure with Deep Neural Networks Fusing WSI and Immunofluorescence Images

10/1/2023
Thesis
Master's thesis on predicting renal failure in nephropathy patients by combining whole-slide renal biopsy images with immunofluorescence images using deep neural networks.

Background

2016 - 2019

University of Modena and Reggio Emilia

B.Sc. in Computer Science

Coursework included data structures and algorithms, databases, object-oriented programming, and statistics.

2020 - 2023

University of Modena and Reggio Emilia

M.Sc. in Computer Science - Artificial Intelligence track

Graduated 110/110 cum laude. Coursework included machine learning, deep learning, computer vision, and large-scale data analysis.

Dec 2023 - present

University of Modena and Reggio Emilia / AImageLab

Research Fellow / Machine Learning Engineer

Develops deep-learning models and pipelines for medical imaging, clinical tabular data, cancer cell analysis, pharmaceutical vial inspection, and a Django-based kidney failure risk prediction application.