Yasutaka Furukawa

Simon Fraser University, Washington University in St. Louis

Papers

Total Citations

341

H-Index

About

Yasutaka Furukawa is a leading figure in computer vision, renowned for his pioneering work in 3D scene understanding and reconstruction from single images. His research primarily focuses on deep learning architectures for geometric perception, particularly in detecting and modeling planar surfaces in complex environments. Furukawa’s most impactful contribution is **PlaneRCNN**, a novel deep neural network that detects and reconstructs piecewise planar regions from a single RGB image. By adapting Mask R-CNN to predict plane parameters and segmentation masks, PlaneRCNN enables accurate 3D plane detection without requiring multiple views or depth sensors—a breakthrough for applications in augmented reality, robotics, and indoor mapping. This work has garnered over 240 citations, underscoring its influence. Additionally, Furukawa introduced the **Scene Agnostic Network (SANet)** for camera localization, a model that generalizes across scenes without per-scene retraining, achieving 91 citations for its efficiency in online applications. His contributions to large-scale 3D modeling of urban indoor and outdoor scenes have further solidified his reputation. Furukawa’s innovations bridge the gap between 2D vision and 3D geometry, making him a key researcher for students and professionals exploring single-image reconstruction and scene-agnostic learning.