Roberto Bigazzi

I am a Postdoctoral research fellow at the AImageLab research laboratory. The main topics of my research are Computer Vision and Deep Learning, in particular on robotic visual navigation.

I've spent a period as Visiting Student Researcher at Stanford University in the Autonomous Systems Lab (ASL) under Prof. Marco Pavone. I did my PhD at University of Modena and Reggio Emilia advised by Prof. Rita Cucchiara, while I pursued my Bachelor's and Master's degrees at Polytechnic University of Milan with an exchange period at Technische Universität Wien.

I am a IEEE (Institute of Electrical and Electronics Engineers) and a CVF (The Computer Vision Foundation) member.

Email  /  CV  /  Resume  /  BC  /  Google Scholar  /  LinkedIn  /  GitHub

In this thesis, I present the research work carried out during my Ph.D. that was focused on the challenges defined by the field of Embodied Artificial Intelligence. I would like to thank again my family, friends, and supervisor for the support during the last three years.

We propose a novel architecture inspired by Generative Adversarial Networks (GANs) that produces meaningful and well-formed synthetic instructions to improve navigation agents’ performance. The validation analysis of our proposal on REVERIE and R2R highlights the promising aspects of our proposal, achieving state-of-the-art results.

We propose a novel approach that combines visual and fine-tuned CLIP features to generate grounded language-visual features for mapping. These region classification features are seamlessly integrated into a global grid map using an exploration-based navigation policy.

We propose and evaluate an approach that combines recent advances in visual robotic exploration and image captioning on images generated through agent-environment interaction. Our approach can generate smart scene descriptions that maximize semantic knowledge of the environment and avoid repetitions.

We use explainable maps to visualize model predictions and highlight the correlation between observed entities and words generated by a captioning model during the embodied exploration.

We propose Spot the Difference: a novel task for Embodied AI where the agent has access to an outdated map of the environment and needs to recover the correct layout in a fixed time budget.

We propose to train a navigation model with a purely intrinsic reward signal to guide exploration, which is based on the impact of the robot's actions on its internal representation of the environment.

We build and release a new 3D space for embodied navigation with unique characteristics: the one of a complete art museum. We name this environment ArtGallery3D (AG3D).

We detail how to transfer the knowledge acquired in simulation into the real world describing the architectural discrepancies that damage the Sim2Real adaptation ability of models trained on the Habitat simulator and we propose a novel solution tailored towards the deployment in real-world scenarios.

We devise an embodied setting in which an agent needs to explore a previously unknown environment while recounting what it sees during the path. The agent needs to navigate the environment driven by an exploration goal, select proper moments for description, and output natural language descriptions of relevant objects and scenes.

Reviewing Committees

Journals:

• IEEE Robotics and Automation Letters (RA‑L)

• IEEE Geoscience and Remote Sensing Letters (GRSL)

• IEEE Pattern Recognition Letters (PRL)

• Transactions on Multimedia Computing Communications and Applications (TOMM)

Conferences:

• IEEE/CVF Computer Vision and Pattern Recognition Conference Workshops (CVPRW)

• European Conference on Computer Vision (ECCV)

• IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)

• IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)

• IEEE International Conference on Robotics and Automation (ICRA)

• IAPR International Conference on Pattern Recognition (ICPR)

• ACM International Conference on Multimedia (ACMMM)

• International Conference on Image Analysis and Processing (ICIAP)

Other:

• ELLIS Ph.D. Program

Certificates and Memberships

• IEEE (Institute of Electrical and Electronics Engineers) Membership

• CVF (The Computer Vision Foundation) Membership

• Visiting Student Researcher @ Stanford University [certificate]

• English Certificate TOEIC (C1) [certificate]

• English Certificate FCE (B2)

• Transcript of Records for Erasmus @ TU Wien [certificate]

• European Computer Driving Licence (ECDL) [certificate]

Teaching Activities

• Computer Architectures - Prof. Rita Cucchiara, Prof. Simone Calderara, 2020-2023

• Project Tutor for Computer Vision and Cognitive Systems - Prof. Rita Cucchiara, Dott. Lorenzo Baraldi, 2019-2023

• Project Tutor for AI for Automotive - Prof. Rita Cucchiara, Dott. Lorenzo Baraldi, 2021-2022

• Python and Machine Learning course for Banco BPM - Prometeia, 2021

• Data Analysis and Data Visualization course - IFOA Bologna, 2021

• Deep Learning, Artificial Intelligence e Neurolinguistic Processing (IBM, SAS) course - IFOA Modena, 2021

• Project Tutor for Neural Network Computing, AI and Machine Learning for Automotive - Prof. Rita Cucchiara, Dott. Lorenzo Baraldi, 2020-2021

Thesis Supervision

• Davide Borghi - Evaluating the Effect of Exploration Bonuses on Deep Reinforcement Learning-Based Agents [pdf]

Courses and Summer Schools

• Advanced Course on Data Science and Machine Learning (ACDL 2021) - Siena (SI), Italy [certificate]

• Interactive Internet of Things e Smart Object Design Course - Polytechnic University of Milan, Italy

International Conferences

• IEEE International Conference on Robotics and Automation (ICRA 2023) - London, UK [certificate]

• IEEE International Conference on Robotics and Automation (ICRA 2022) - Philadelphia, US [certificate]

• International Conference on Computer Analysis of Images and Patterns (CAIP 2021) - Nicosia, Cyprus (Remote) [certificate]

• IAPR International Conference on Pattern Recognition (ICPR 2020) - Milan, Italy (Remote)

For any issues with this page please let me know here
* thanks to Jon Barron for the codebase of this website.