A Fast Hybrid Cascade Network for Voxel-based 3D Object Classification

TL;DR

This paper introduces a hybrid cascade network for voxel-based 3D object classification that improves accuracy and speed by handling models of varying difficulty and incorporating signed distance values for voxels.

Contribution

It proposes a novel three-stage cascade architecture that balances accuracy and efficiency, and leverages signed distance voxel encoding for better classification performance.

Findings

Significant accuracy improvement with signed distance voxel encoding

Hugely faster inference times compared to state-of-the-art methods

Effective handling of easy, moderate, and hard 3D models

Abstract

Voxel-based 3D object classification has been thoroughly studied in recent years. Most previous methods convert the classic 2D convolution into a 3D form that will be further applied to objects with binary voxel representation for classification. However, the binary voxel representation is not very effective for 3D convolution in many cases. In this paper, we propose a hybrid cascade architecture for voxel-based 3D object classification. It consists of three stages composed of fully connected and convolutional layers, dealing with easy, moderate, and hard 3D models respectively. Both accuracy and speed can be balanced in our proposed method. By giving each voxel a signed distance value, an obvious gain regarding the accuracy can be observed. Besides, the mean inference time can be speeded up hugely compared with the state-of-the-art point cloud and voxel based methods.