CarveNet: Carving Point-Block for Complex 3D Shape Completion

TL;DR

CarveNet introduces a novel 3D shape completion method using point-block carving guided by a new network architecture and data augmentation, effectively handling complex shapes and diverse partial point clouds.

Contribution

The paper proposes Point-block Carving (PC) and CarveNet, a new network architecture with exclusive convolution, along with SensorAug data augmentation, advancing 3D point cloud completion capabilities.

Findings

Outperforms state-of-the-art methods on ShapeNet and KITTI datasets.

Effectively handles complex 3D shapes with high curvature and hollowed-out structures.

Demonstrates robustness to diverse partial point cloud patterns.

Abstract

3D point cloud completion is very challenging because it heavily relies on the accurate understanding of the complex 3D shapes (e.g., high-curvature, concave/convex, and hollowed-out 3D shapes) and the unknown & diverse patterns of the partially available point clouds. In this paper, we propose a novel solution,i.e., Point-block Carving (PC), for completing the complex 3D point cloud completion. Given the partial point cloud as the guidance, we carve a3D block that contains the uniformly distributed 3D points, yielding the entire point cloud. To achieve PC, we propose a new network architecture, i.e., CarveNet. This network conducts the exclusive convolution on each point of the block, where the convolutional kernels are trained on the 3D shape data. CarveNet determines which point should be carved, for effectively recovering the details of the complete shapes. Furthermore, we propose a sensor-aware method for data augmentation,i.e., SensorAug, for training CarveNet on richer patterns of partial point clouds, thus enhancing the completion power of the network. The extensive evaluations on the ShapeNet and KITTI datasets demonstrate the generality of our approach on the partial point clouds with diverse patterns. On these datasets, CarveNet successfully outperforms the state-of-the-art methods.