3D Reconstruction in the Era of Foundation Models

Speaker:  Prof. Matteo Poggi

April 17^th, 2026 | 2:00 pm
DEIB, Seminar Room "N. Schiavoni" (Bld. 20A)

Contact:  Prof. Federica Arrigoni

Abstract

On April 17^th, 2026, at 2:00 pm the seminar on "3D Reconstruction in the Era of Foundation Models" will take place in DEIB Seminar Room "Nicola Schiavoni" (Building 20A).

The past year has witnessed a paradigm shift in 3D reconstruction, fuelled by the emergence of the first 3D foundation models. While these models have unlocked unprecedented opportunities, they have raised the bar in terms of hardware requirements and access barriers to the field. Training these architectures from scratch is no longer a standard task; it is a resource-intensive challenge requiring dozens of GPUs and vast datasets. How do we still innovate when these resources are out of reach? In this talk, we challenge the necessity of "starting from zero” and repurposing pre-trained foundation models to design a new generation of architectures for depth estimation and related tasks. This allows for leveraging existing data and computing efforts to achieve state-of-the-art, without the need for those resources ourselves.
This talk is part of the Imaging Seminar from EMJM in Imaging.

Short Bio

Matteo Poggi is an Associate Professor with Alma Mater Studiorum, University of Bologna. His research interests span in the area of 3D Computer Vision, ranging from low-level depth estimation, optical flow estimation, online 3D reconstruction and domain adaptation. 
He has been serving as an Associate Editor for the International Journal of Computer Vision (IJCV) and as Area Chair for ICRA 2026 and CVPR 2026, as well as regularly serves as a reviewer for CVPR/ICCV/ECCV/BMVC conferences, being acknowledged as an outstanding reviewer 10+ times.