Credits and usage
Research material
The abstract, experiment figures, robot photographs, reconstruction visualizations, and supplementary video come from the AnchorHEC manuscript and its supporting materials. Tables II and IV are rendered directly from the manuscript PDF, including their captions and footnotes.
Interactive scenes
The 3D views use retained point clouds and camera poses from the figure-authoring pipeline. Physical scenes retain Figure 5’s saved point IDs and filters, with no additional subsampling or filling; camera estimates are not recomputed. Before shows the native reconstruction; After applies the stored scale and robot-base alignment. The cloud is not dense geometry optimized by bundle adjustment. Each scene is an individual example.
The locally hosted viewer is a locally adapted distribution of Viser 1.0.24, licensed under Apache License 2.0. It reads static scene recordings in the browser. A small local adapter links image selection and before/after controls to Viser’s scene state. The upstream renderer is retained.
Website layout
The layout follows the academic project-page format of Nerfies by Park et al. The page HTML and CSS were rewritten for AnchorHEC; the Nerfies template, fonts, and analytics were not copied. No analytics are included.
Third-party footage
Introductory footage in the overview video is excerpted and arranged from the DROID dataset, Khazatsky et al., DROID: A Large-Scale In-the-Wild Robot Manipulation Dataset, RSS 2024, under Creative Commons Attribution 4.0. The footage has been edited for this introduction.
Usage
The manuscript, media, and third-party assets retain their respective rights. This website does not grant additional rights to the underlying research code or datasets. Future code or data releases will include their own licensing and attribution information.