Sankeerth Durvasula

Papers

Total Citations

H-Index

About

Sankeerth Durvasula is a researcher at the forefront of real-time perception systems for robotics and autonomous platforms. His work bridges the critical gap between high-speed sensor data and efficient on-device computation, with a primary focus on 3D mapping and event-based vision. Durvasula’s key contribution, "VoxelCache" (2022), addresses the fundamental challenge of real-time 3D mapping by enabling continuous fusion of depth data from sensors in phones, robots, and autonomous vehicles into a single, live 3D model—a capability essential for AR/VR and robotic navigation. Building on this, his work "Ev-Conv: Fast CNN Inference on Event Camera Inputs for High-Speed Robot Perception" (2023) pioneers efficient neural network inference on event camera data. By leveraging the microsecond-level temporal resolution of event cameras, this research unlocks robust perception in rapidly changing, high-dynamic-range environments, enabling robots to "see" and react at unprecedented speeds. Though early in his career, Durvasula’s contributions are already shaping how autonomous systems process visual information in real time, laying the groundwork for faster, more responsive robots and immersive 3D experiences.