Michael Firman

Papers

Total Citations

H-Index

About

Michael Firman’s research lies at the intersection of 3D computer vision, augmented reality, and efficient scene understanding. His major contributions center on developing practical, real-time methods for reconstructing and interpreting 3D environments from posed RGB images—a critical capability for AR and robotics. His 2023 work on “Heightfields for Efficient Scene Reconstruction for AR” (5 citations) introduces a novel, lightweight representation that enables fast and accurate 3D scene reconstruction, directly addressing the computational bottlenecks of traditional depth-fusion approaches. In 2024, his paper “AirPlanes: Accurate Plane Estimation via 3D-Consistent Embeddings” (3 citations) tackles the fundamental problem of planar surface extraction from images, demonstrating that a surprisingly strong baseline can be built by combining popular clustering techniques with learned 3D-consistent embeddings. This work provides a robust, data-driven solution for downstream tasks like mapping and object interaction. While still early in his career, Firman’s focus on bridging learning-based methods with real-time performance positions him as a rising contributor to practical AR and robotics systems, with his work already influencing how researchers approach efficient 3D scene parsing.