P2C: Self-Supervised Point Cloud Completion from Single Partial Clouds

TL;DR

P2C tackles point cloud completion under a strict self-supervised setting by using only a single partial observation per object. It introduces a patch-based pipeline that partitions input patches into observed, completed, and latent groups, coupled with four losses, including Region-Aware Chamfer Distance and Normal Consistency Constraint, to infer complete shapes without ground-truth completions. The key contributions are the RCD distance that focuses on regions around skeleton points and the NCC that enforces local planarity, enabling high-quality completions comparable to methods trained with complete shapes and outperforming those learned from multiple partial observations. The approach demonstrates strong results on ShapeNet-derived synthetic data and real-world ScanNet scans, highlighting practical impact for scalable 3D shape completion in realistic, data-scarce scenarios.

Abstract

Point cloud completion aims to recover the complete shape based on a partial observation. Existing methods require either complete point clouds or multiple partial observations of the same object for learning. In contrast to previous approaches, we present Partial2Complete (P2C), the first self-supervised framework that completes point cloud objects using training samples consisting of only a single incomplete point cloud per object. Specifically, our framework groups incomplete point clouds into local patches as input and predicts masked patches by learning prior information from different partial objects. We also propose Region-Aware Chamfer Distance to regularize shape mismatch without limiting completion capability, and devise the Normal Consistency Constraint to incorporate a local planarity assumption, encouraging the recovered shape surface to be continuous and complete. In this way, P2C no longer needs multiple observations or complete point clouds as ground truth. Instead, structural cues are learned from a category-specific dataset to complete partial point clouds of objects. We demonstrate the effectiveness of our approach on both synthetic ShapeNet data and real-world ScanNet data, showing that P2C produces comparable results to methods trained with complete shapes, and outperforms methods learned with multiple partial observations. Code is available at https://github.com/CuiRuikai/Partial2Complete.

Figures (7)

• Figure 1: Conceptual comparison of point cloud completion schemes. Let $x_{k}^{(i)}$ be the $k$-th incomplete observation of object $i$, while $\hat{y}^{(i)}$ and $y^{(i)}$ be the corresponding completed prediction and ground truth, respectively. (a) Supervised approaches rely on paired partial-complete samples. (b) Unpaired methods require partial point clouds and complete examples to guide predictions to match the input shape and follow the complete shape distribution. (c) Weakly-supervised models learn completion based on consistency across multi-view partial samples of an object. (d) Our scheme differs from existing settings as only a single partial observation per object instance is available for learning.
• Figure 2: The Pipeline of P2C. Starting from the partial point cloud $P_p$, we divide it into patches and partition these patches into three groups ($G_{rec}, G_{com}, G_{latent}$). The encoder takes $G_{rec}$ to produce features $f$ then the decoder generates a predicted point cloud $P_c$ based on $f$. $G_{latent}$ is never observed by the encoder, we resample corresponding regions $G_{latent}^{'}$ in $P_c$ to yield another feature embedding $f'$. The overall loss has four components. The reconstruction loss $\mathcal{L}_{r}$ and completion loss $\mathcal{L}_{c}$ are realized by RCD. The Latent reconstruction loss $\mathcal{L}_{f}$ and the normal consistency constraint $\mathcal{L}_{ncc}$ are introduced to regularize the inference.
• Figure 3: Comparison of pulling direction to minimize different distance measures. (a) CD takes the nearest neighbor for every point in the predicted set, leading to restrictions in completing missing parts; (b) UCD considers the nearest neighbor only for every point in the target set, resulting in no moving directions for noisy points; (c) RCD is aware of observed and unseen regions and thus only evaluates point distance for observed regions, pulling outlier points to the underlying surface while allowing completion of unseen parts.
• Figure 4: Illustration of the effect of NCC in a 2D case: The variance of normal similarity is lower when the point follows the underlying surface, as shown in (a) and (c), while the variance is larger when the new point results in a surface that diverges from the existing surface curvature, as shown in (b) and (d).
• Figure 5: Visual comparison of point cloud completion results on the 3D-EPN dataset.