Event

Authors Release MILD Dataset and AprilVINS Localization Method

1 reports · 1 sources

Junjie Zhang and other authors released the MILD manipulation-interface localization dataset and the AprilVINS visual-inertial localization method, aimed at robot end-effector localization in UMI-style manipulation demonstrations. The real subset contains 86 sensor sequences collected with Insta360 X5 and Insight9, covering 15 repeated tabletop tasks, and provides calibration assets and end-effector reference trajectories for each execution; the simulated subset MILD-Sim extends task coverage in Isaac Sim for controlled manipulation replay studies. The authors state that the code, dataset, and evaluation checklist will be released after the paper is accepted and are not yet publicly available.