Daniel Lenton

Dyson (United Kingdom), Imperial College London

Papers

Total Citations

121

H-Index

About

Daniel Lenton is a researcher working at the intersection of computer vision, robotics, and deep learning infrastructure. His most prominent contribution is **MoreFusion** (2020), a system for multi-object 6D pose estimation using volumetric fusion, which has accumulated over 100 citations and represents a significant advance in how robots and smart devices construct object-aware scene representations. By combining recognized object models with non-parametric reconstructions of unrecognized structures, MoreFusion enables more robust reasoning about contact, physics, and occlusion — capabilities essential for real-world robotic manipulation and autonomous systems. Beyond perception, Lenton has also contributed to the foundations of machine learning development through **Ivy** (2021), a templated deep learning framework designed to abstract and unify existing frameworks such as TensorFlow, PyTorch, and JAX. By standardizing function signatures and input-output behavior across platforms, Ivy addresses the longstanding challenge of inter-framework portability, empowering researchers and engineers to write framework-agnostic code. Together, these works reflect Lenton's broader ambition to make intelligent systems — both in perception and in the tools that build them — more flexible, interoperable, and capable of operating effectively in complex, real-world environments.