Search papers, labs, and topics across Lattice.
This paper introduces SAM-H and WOFTSAM, two novel planar trackers leveraging SAM 2 for robust segmentation tracking and 8-DoF homography pose estimation. SAM-H achieves robustness to appearance changes by estimating homographies from segmentation mask contours, while WOFTSAM enhances the WOFT tracker by incorporating SAM-H's re-detection capabilities. Experimental results on POT-210 and PlanarTrack benchmarks demonstrate state-of-the-art performance, with WOFTSAM achieving significant gains on PlanarTrack, alongside improved ground-truth annotations for more precise benchmarking.
By combining SAM's segmentation prowess with homography estimation, these new planar trackers leapfrog existing methods by a significant margin, setting a new state-of-the-art on standard benchmarks.
We present SAM-H and WOFTSAM, novel planar trackers that combine robust long-term segmentation tracking provided by SAM 2 with 8 degrees-of-freedom homography pose estimation. SAM-H estimates homographies from segmentation mask contours and is thus highly robust to target appearance changes. WOFTSAM significantly improves the current state-of-the-art planar tracker WOFT by exploiting lost target re-detection provided by SAM-H. The proposed methods are evaluated on POT-210 and PlanarTrack tracking benchmarks, setting the new state-of-the-art performance on both. On the latter, they outperform the second best by a large margin, +12.4 and +15.2pp on the p@15 metric. We also present improved ground-truth annotations of initial PlanarTrack poses, enabling more accurate benchmarking in the high-precision p@5 metric. The code and the re-annotations are available at https://github.com/serycjon/WOFTSAM
