Learning 3D Shapes as Multi-Layered Height-maps using 2D Convolutional Networks

TL;DR

This paper introduces a new 3D shape representation using multi-layered height-maps that enables the use of 2D CNNs, achieving state-of-the-art classification results efficiently.

Contribution

The authors propose a multi-layered height-map representation for 3D shapes and a view merging method, allowing effective use of 2D CNNs for 3D shape classification.

Findings

Achieved state-of-the-art classification accuracy on ModelNet dataset.

Memory-efficient input representation compared to voxel-based methods.

Effective multi-view merging technique for 3D shape descriptors.

Abstract

We present a novel global representation of 3D shapes, suitable for the application of 2D CNNs. We represent 3D shapes as multi-layered height-maps (MLH) where at each grid location, we store multiple instances of height maps, thereby representing 3D shape detail that is hidden behind several layers of occlusion. We provide a novel view merging method for combining view dependent information (Eg. MLH descriptors) from multiple views. Because of the ability of using 2D CNNs, our method is highly memory efficient in terms of input resolution compared to the voxel based input. Together with MLH descriptors and our multi view merging, we achieve the state-of-the-art result in classification on ModelNet dataset.