Gabriele Berton

Gabriele Berton is a postdoctoral researcher at Amazon, focusing on Multimodal Large Language Models and advanced computer vision applications. He holds a PhD in Computer Vision from the Polytechnic of Turin under Prof. Barbara Caputo. His research expertise includes visual geo-localization, place recognition, and scalable image retrieval systems. He previously worked at the Italian Institute of Technology (IIT) and was a visiting scholar at Carnegie Mellon University's AirLab under Prof. Sebastian Scherer. His work on EarthLoc is utilized by NASA for localizing International Space Station imagery.

Education:

• PhD in Computer Vision, Polytechnic of Turin (2022)
• Master’s/Pre-PhD work at Italian Institute of Technology

Research Focus:

Berton’s work emphasizes robust visual localization systems, including 3D mesh-based place recognition (MeshVPR), astronaut photography indexing (EarthLoc), and iterative coregistration techniques (EarthMatch). He develops scalable algorithms for large-scale applications, balancing performance with computational efficiency.

Key Achievements:

• First-author papers at CVPR, ICCV, ECCV (7+)
• First Prize in Best Student Research at EuroCon 2023
• Co-developed CosPlace framework for large-scale visual localization

Labs/Teams:

Collaborates with AirLab at CMU and MaLGa Lab. Active contributor to the RoboTo robotics team, focusing on computer vision for robotics competitions.