Mengtian Li

Papers

Total Citations

H-Index

About

Mengtian Li is a researcher at the forefront of efficient computer vision for autonomous systems, with a primary focus on enabling robust perception under severe computational and resource constraints. His work directly addresses the critical safety and latency challenges in autonomous navigation and robotics. Li’s most impactful contribution, "FOVEA: Foveated Image Magnification for Autonomous Navigation" (2021, 27 citations), introduces a novel, biologically-inspired approach that intelligently allocates processing power to image regions based on their importance, moving beyond naive downsampling to significantly boost object detector performance without sacrificing real-time speed. He further advances the field of model compression with his work on "Learning Lightweight Object Detectors via Multi-Teacher Progressive Distillation" (2023), tackling the difficult problem of distilling knowledge from multiple, complex teacher models into a single, efficient student detector for edge deployment. By pioneering methods that bridge the gap between high-accuracy perception and the strict memory and compute budgets of real-world robots and edge devices, Li is shaping the future of practical, deployable AI.