Reproduce “Towards Accurate State Estimation: Motion Dynamics Kalman Filter for 3D Multi-Object Tracking” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Mohamed Nagy, Naoufel Werghi, Bilal Hassan, Jorge Dias, Majid Khonji
MD-KF adapts Kalman-filter predictions using each object’s recent motion, adjusting velocity, acceleration, and jerk rather than maintaining parallel models. On KITTI validation with VirConv detections and a simulated 20-frame detection gap, it scores 83.820 HOTA, a composite detection-and-association score, versus 83.461 for IMM-KF. Its reported mean throughput is 1.72× IMM-KF’s, but IMM-KF slightly leads in the 10-frame test.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.