Depth Completion Using Geometry-Aware Embedding

Exploiting internal spatial geometric constraints of sparse LiDARs is beneficial to depth completion, however, has been not explored well. This paper proposes an efficient method to learn geometry-aware embedding, which encodes the local and global geometric structure information from 3D points, e. g. , scene layout, object's sizes and shapes, to guide dense depth estimation. Specifically, we utilize the dynamic graph representation to model generalized geometric relationship from irregular point clouds in a flexible and efficient manner. Further, we joint this embedding and corresponded RGB appearance information to infer missing depths of the scene with well structure-preserved details. The key to our method is to integrate implicit 3D geometric representation into a 2D learning architecture, which leads to a better trade-off between the performance and efficiency. Extensive experiments demonstrate that the proposed method outperforms previous works and could reconstruct fine depths with crisp boundaries in regions that are over-smoothed by them. The ablation study gives more insights into our method that could achieve significant gains with a simple design, while having better generalization capability and stability. The code is available at https://github.com/Wenchao-Du/GAENet.

Authors

• Wenchao Du
• Hu Chen 0002
• Hongyu Yang
• Yi Zhang 0018

Keywords

• Point cloud compression
• Three-dimensional displays
• Laser radar
• Codes
• Automation
• Shape
• Layout
• Depth Completion
• Local Information
• Point Cloud
• Global Information
• Geometric Structure
• Global Structure
• Generalization Capability
• 3D Point
• Geometric Constraints
• Geometric Information
• Geometric Relationship
• Depth Estimation
• Dynamic Representation
• Geometric Representation
• Dynamic Graph
• Appearance Information
• Local Structure Information
• Dense Depth
• Local Geometric Structure
• Global Structural Information
• KITTI Dataset
• Outdoor Scenes
• Mean Absolute Error
• Convolutional Neural Network
• RGB Images
• Ability Of The Model
• Visual Features
• Feature Maps
• Feature Points
• 3D Space