ComPC: Completing a 3D Point Cloud with 2D Diffusion Priors
This addresses the issue of self-occlusion in point clouds for applications like robotics or AR/VR, offering a zero-shot approach that works across unseen object categories, though it builds incrementally on prior diffusion and Gaussian splatting techniques.
The paper tackles the problem of completing incomplete 3D point clouds from sensors by proposing a test-time framework that uses 2D diffusion priors without training, outperforming existing methods on synthetic and real-world data.
3D point clouds directly collected from objects through sensors are often incomplete due to self-occlusion. Conventional methods for completing these partial point clouds rely on manually organized training sets and are usually limited to object categories seen during training. In this work, we propose a test-time framework for completing partial point clouds across unseen categories without any requirement for training. Leveraging point rendering via Gaussian Splatting, we develop techniques of Partial Gaussian Initialization, Zero-shot Fractal Completion, and Point Cloud Extraction that utilize priors from pre-trained 2D diffusion models to infer missing regions and extract uniform completed point clouds. Experimental results on both synthetic and real-world scanned point clouds demonstrate that our approach outperforms existing methods in completing a variety of objects. Our project page is at \url{https://tianxinhuang.github.io/projects/ComPC/}.