AI & Computer Vision Research Hub

Led by Dr. Silvia Cascianelli

Welcome to our AI & Computer Vision research hub. We focus on exploring cutting-edge techniques in image generation, handwriting imitation, and document understanding, while fostering an environment that emphasizes mentorship and collaboration. Below, you’ll find highlights of our recent projects and the brilliant minds collaborating on them.

Table of Contents

1. Introduction
2. Recent Projects
   • Image Generation
   • Handwriting Imitation
   • Document Understanding
3. Our Team

Introduction

We are a group of researchers delving into challenges at the forefront of AI. We tackle complex tasks in AI-driven content creation, strive to replicate distinctive handwriting styles, and unlock insights from historical and modern documents.

Recent Projects

Image Generation

We have developed efficient and lightweight methods to generate images with desired characteristics. Our work particularly explores how Diffusion Models can be adapted at inference time to produce semantically coherent or uniquely formatted images.

• Merging and Splitting Diffusion Paths for Semantically Coherent Panoramas
  [Fabio Quattrini, Vittorio Pippi, Silvia Cascianelli, Rita Cucchiara]
  arXiv:2408.15660

• Alfie: Democratising RGBA Image Generation With No $$$
  [Fabio Quattrini, Vittorio Pippi, Silvia Cascianelli, Rita Cucchiara]
  arXiv:2408.14826

Handwriting Imitation

Our research explores algorithms to generate images of text in various handwriting styles, how to evaluate the generated results, and how these findings can be applied in real-world contexts (e.g., personalized text rendering).

• VATr++: Choose Your Words Wisely for Handwritten Text Generation
  [Bram Vanherle, Vittorio Pippi, Silvia Cascianelli, Nick Michiels, Frank Von Reeth, Rita Cucchiara]
  arXiv:2402.10798

• Handwritten Text Generation from Visual Archetypes
  [Vittorio Pippi, Silvia Cascianelli, Rita Cucchiara]
  arXiv:2303.15269

• HWD: A Novel Evaluation Score for Styled Handwritten Text Generation
  [Vittorio Pippi, Fabio Quattrini, Silvia Cascianelli, Rita Cucchiara]
  arXiv:2310.20316

• How to Choose Pretrained HTR Models for Single Writer Fine-Tuning
  [Vittorio Pippi, Silvia Cascianelli, Christopher Kermorvant, Rita Cucchiara]
  arXiv:2305.02593

Document Understanding

From analyzing complex 2D/3D scans of historical papyri to modern ESG reports, our team builds systems to detect ink traces, parse text, interpret layout, and uncover valuable writer-centric details.

• μgat: Improving Single-Page Document Parsing by Providing Multi-Page Context
  [Carmine Zaccagnino, Fabio Quattrini, Silvia Cascianelli, Laura Righi, Rita Cucchiara]
  arXiv:2408.15646

• Volumetric FFC for Detecting Ink on the Carbonized Herculaneum Papyri
  [Fabio Quattrini, Vittorio Pippi, Silvia Cascianelli, Rita Cucchiara]
  arXiv:2308.05070

• Evaluating Synthetic Pre-Training for Handwriting Processing Tasks
  [Vittorio Pippi, Silvia Cascianelli, Lorenzo Baraldi, Rita Cucchiara]
  arXiv:2304.01842

• The LAM Dataset: A Novel Benchmark for Line-Level Handwritten Text Recognition
  [Silvia Cascianelli, Vittorio Pippi, Martin Maarand, Marcella Cornia, Lorenzo Baraldi, Christopher Kermorvant, Rita Cucchiara]
  arXiv:2208.08109

• Boosting Modern and Historical HTR with Deformable Convolutions
  [Silvia Cascianelli, Marcella Cornia, Lorenzo Baraldi, Rita Cucchiara]
  arXiv:2305.02593

Our Team

A significant part of our success comes from a dedicated group of researchers, interns, and students:

• Silvia Cascianelli: Assistant Professor.
  Lorem, ipsum, and dolor.

• Vittorio Pippi: PhD student (National PhD in AI); recently interned at Amazon Research Berlin.
  Grounded, generous, and sharp.

• Fabio Quattrini: PhD student at UniMoRe; also interning at Amazon Research Berlin.
  Creative, kind, and direct.

• Carmine Zaccagnino: Recent MSc graduate and current research intern at UniMoRe.
  Hardworking, proactive, and passionate.

• Pau Torras Coloma: PhD student at CVC-UAB; spent 1.5 months at UniMoRe as a visiting intern.
  Committed, curious, and adaptable.

• Bram Vanherle: CV Engineer at Colruyt Group Smart Innovation; spent 5 months of internship at UniMoRe.
  Concrete, easygoing, and consistently solid.

Contact

Thank you for your interest! If you find our work compelling or would like to discuss potential collaborations, please reach out:

• Email: scascianelli[at]unimore[dot]it
• LinkedIn: linkedin.com/in/silvia-cascianelli-330760104

We look forward to hearing from you!