Carlos Esteves

University of Pennsylvania, Google (United States)

Papers

Total Citations

H-Index

About

Carlos Esteves is a leading researcher at the intersection of computer vision, robotics, and geometric deep learning, with a focus on rotation estimation and 3D pose prediction. His work addresses fundamental challenges in modeling symmetries and uncertainties on non-Euclidean manifolds. In his highly cited 2020 paper, "An Analysis of SVD for Deep Rotation Estimation" (32 citations), Esteves provided a rigorous theoretical and empirical analysis of using Singular Value Decomposition (SVD) to enforce orthogonal and special orthogonal constraints in neural networks, establishing a standard approach for deep rotation regression. He further advanced the field with "Implicit-PDF: Non-Parametric Representation of Probability Distributions on the Rotation Manifold" (7 citations), which introduced a novel framework for capturing multi-modal pose distributions—critical for handling symmetric objects and ambiguous viewpoints in single-image pose estimation. Esteves’s work is notable for its mathematical depth and practical impact, bridging differentiable geometry with robust deep learning pipelines. His contributions are widely adopted in robotics, augmented reality, and autonomous systems, where precise and uncertainty-aware rotation estimation is essential.