6DOPE-GS: Online 6D Object Pose Estimation using Gaussian Splatting
Yufeng Jin, Vignesh Prasad, Snehal Jauhri, Mathias Franzius, Georgia Chalvatzaki
We introduce 6DOPE-GS, a method for online 6D object pose estimation and tracking. We leverage fast differentiable rendering via 2D Gaussian Splatting with dynamic keyframe selection and opacity-based pruning to jointly optimize object pose and 3D reconstruction while ensuring high spatial coverage of the object and adaptive Gaussian control to reduce computational load.
2HandedAfforder: Learning Precise Actionable Bimanual Affordances from Human Videos
Marvin Heidinger*, Snehal Jauhri*, Vignesh Prasad, Georgia Chalvatzaki
We propose (i) a framework for extracting bimanual affordance data from human activity video datasets and (ii) a novel VLM-based bimanual affordance prediction model, that predicts actionable bimanual affordance regions from task-related text prompts.