2 articles
We built and operate an autonomous navigation and data-capture skill end-to-end on a Unitree GO2, an NVIDIA Jetson, and a Livox MID-360. This is the on-robot side of Datameister's Physical AI work, designed to feed the manipulation skills coming next.
Read ArticleClean, anatomy-aware retopology remains a high-stakes bottleneck that existing manual, semi-automatic, and fully automatic tools only partially address. Retopomeister identifies anatomical anchor points, selects appropriate curated source topologies, and uses neural wrapping to conform that topology onto the target mesh with symmetry preservation, while allowing artists to adjust anchors and maintain creative control. By transferring production-ready metadata like UV unwraps, vertex groups, and material seams along with the topology, it reduces downstream work across the entire pipeline. The prototype system showcases promising directions: enhanced keypoints, interactive loop sketching, and a more iterative workflow.
Read Article
