Felix Leeb

Papers

Total Citations

H-Index

About

Felix Leeb’s research lies at the intersection of deep learning, robotics, and visuomotor control, with a focus on enabling machines to perceive and interact with dynamic environments. His most influential work introduces **SE3-Pose-Nets**, a structured deep dynamics model that learns low-dimensional pose embeddings from visual input for planning and control. By leveraging an encoder-decoder architecture, Leeb’s model captures the spatial transformations of objects in 3D space, allowing robots to reason about motion and manipulation tasks more efficiently than unstructured approaches. This work has garnered over 75 combined citations, reflecting its impact on the field of model-based reinforcement learning and robotic perception. Leeb further extended his contributions with **Motion-Nets**, a method for 6D tracking of unknown objects in unseen environments using only RGB data. By integrating segmentation with separate translation and rotation models, Motion-Nets bridges pose estimation and tracking, enabling robust object following in novel settings. Together, these works demonstrate Leeb’s commitment to building structured, generalizable models that advance autonomous systems toward more adaptive and intelligent behavior.