AnchorComplete reconstructs full 3D scene meshes from partial scans
AnchorComplete: Recovering Complete Scene Meshes from Partial Point Clouds
Computer Vision and Pattern Recognition
Summary
When 3D scanners look at a scene, they often see just parts of objects, leaving holes and missing bits in the 3D picture. The authors developed AnchorComplete, a way to fill in these missing parts and create a complete 3D model from only what is partly scanned. It does this by combining detailed clues from the scanned parts with a step-by-step process that builds the full scene accurately and keeps everything in the right scale. Tests show it works better than other methods, especially when only a few views are available.
What this means in practice
- •For robotics engineers: Create complete 3D environment maps from limited sensor data to improve robot navigation and interaction.
- •For virtual reality developers: Generate full scene meshes from sparse input to build immersive virtual environments with fewer scans.
Authors
Weirong Chen, Ismail El-Yamany, Ganlin Zhang, Chuanxia Zheng, Andrea Vedaldi, Daniel Cremers
Abstract
Most image-based 3D reconstructors primarily recover visible surfaces, producing partial point clouds with large holes and leaving unobserved geometry unresolved. To address this limitation, we introduce AnchorComplete, a framework that reconstructs complete scene meshes from partial point clouds while preserving observed geometry and supporting scenes of varying scales. Building on TRELLIS.2, our method combines partial-point conditioning, which integrates global scene tokens with local spatial evidence to anchor generation to the input, and metric-scale coarse-to-fine generation, which progressively refines scene geometry within fixed-size windows while maintaining global structural consistency. Experiments on Replica and SCRREAM demonstrate that AnchorComplete consistently outperforms strong baselines in complete-scene reconstruction, particularly under sparse-view settings.