SMF-VO: Direct Ego-Motion Estimation via Sparse Motion Fields

On November 13, 2025, the introduction of Sparse Motion Field Visual Odometry (SMF-VO) revolutionized the field of visual odometry by presenting a lightweight, motion-centric framework that directly estimates instantaneous linear and angular velocity from sparse optical flow. This approach eliminates the need for traditional pose estimation and expensive landmark tracking, which are often computationally demanding. SMF-VO has demonstrated superior efficiency and competitive accuracy on benchmark datasets, achieving over 100 FPS on a Raspberry Pi 5 using only a CPU. This remarkable performance makes it highly suitable for applications in mobile robotics and wearable devices, addressing the limitations of conventional visual odometry and visual inertial odometry methods that rely heavily on large-scale landmark maintenance and continuous map optimization. The development of SMF-VO not only enhances real-time performance but also establishes a scalable and efficient alternative, paving th…