Iro Armeni - Rectified Point Flow: Generic Point Cloud Pose Estimation

Abstract: Robotic systems frequently confront geometric alignment problems such as point cloud registration and multi-part object assembly, which are typically addressed with task-specific pipelines and explicit pose regression. In this talk, I will present Rectified Point Flow, a unified formulation that casts both problems as a single conditional generative task by learning a continuous point-wise velocity field that transports unposed points to their target locations. This approach naturally recovers part poses and intrinsically captures object symmetries without supervision, outperforming prior methods across six benchmarks. I will conclude by discussing applications in cultural heritage, where robust alignment and assembly are critical for documenting, restoring, and reassembling fragmented artifacts and monuments. Bio: Iro Armeni is an assistant professor of Civil and Environmental Engineering at Stanford University. She works at the intersection of civil engineering, architecture, and machine perception to design and construct data-driven sustainable and adaptive environments across the physical and digital space. Iro completed her PhD at Stanford University, Civil and Environmental Engineering Department, with a PhD minor at the Computer Science Department. Afterwards she was a Postdoctoral Fellow at ETH Zurich working at both the Computer Science and Civil, Environmental, and Geomatic Engineering Departments (2023). Prior to her PhD, she received an MSc in Computer Science, an MEng in Architecture and Digital Design, and a Diploma in Architectural Engineering.