Papers
3
Total Citations
499
H-Index
3
About
Guilherme V. Cavalheiro is a leading researcher in computer vision and robotics, with a primary focus on 3D perception, depth estimation, and visual-inertial navigation. His most impactful contribution is the pioneering work on self-supervised depth completion, as demonstrated in his highly cited 2019 paper "Self-Supervised Sparse-to-Dense," which has garnered over 470 citations. This work revolutionized the field by enabling dense depth map estimation from sparse LiDAR and monocular camera inputs without requiring ground-truth labels, directly addressing critical challenges in autonomous driving and robotics such as irregularly spaced depth patterns. Cavalheiro’s approach significantly improved the robustness and practicality of depth sensing in real-world environments. More recently, he has advanced the state of the art in visual-inertial navigation systems (VINS) by integrating Neural Radiance Fields (NeRF) for camera pose regression and uncertainty quantification, as detailed in his 2024 paper "NVINS." This work tackles the computational and quality challenges of NeRF for real-time applications, pushing the boundaries of reliable localization and mapping. With a career marked by high-impact, practical solutions to core perception problems, Cavalheiro continues to shape the future of autonomous systems.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3