Niki Loppi

Papers

Total Citations

H-Index

About

Niki Loppi is a computer vision researcher whose work bridges the gap between robust geometric estimation and deep learning for 3D scene understanding. His primary research areas include stereo matching, visual-inertial odometry (VIO), and domain generalization in perception systems. Loppi’s most notable contribution is the introduction of visual hints expansion for stereo matching, a technique that leverages the sparse, unevenly distributed feature points characteristic of VIO systems to guide and improve generalization in dense depth estimation. This work, published in 2023, addresses a critical challenge in robotics and autonomous navigation: the ability to maintain accurate depth perception across unseen environments without retraining. While his most-cited paper currently holds 4 citations, the work represents a novel conceptual bridge between traditional geometric methods and modern learning-based approaches. Loppi’s research is particularly significant for applications in autonomous driving, drone navigation, and augmented reality, where robust performance across diverse conditions is paramount. His approach offers a practical pathway for making deep stereo networks more resilient to domain shifts, a key bottleneck in real-world deployment of vision systems.