Jonas Koehler

Papers

Total Citations

H-Index

About

Jonas Koehler is a leading researcher in geometric deep learning, with a primary focus on developing neural network architectures for non-Euclidean domains. His most influential work, "Spherical CNNs" (2018), introduced a groundbreaking framework for applying convolutional neural networks to spherical data—a critical need for applications in omnidirectional vision, drone navigation, and autonomous robotics. By redefining convolution on the sphere using the rotation group SO(3), Koehler enabled efficient and equivariant feature learning on spherical signals, overcoming the limitations of traditional planar CNNs. This seminal paper has garnered 38 citations and laid the foundation for subsequent advances in rotation-invariant learning and 3D perception. Koehler’s contributions are particularly notable for bridging theoretical rigor with practical deployment, inspiring a new generation of models for 360-degree imagery and geospatial analysis. His work continues to shape the field, offering powerful tools for researchers tackling problems where data naturally lives on curved surfaces.