Table 12.

Methodology and Limitation(s): 3D Voxel Grid-based 3DOR Methods.

	3D Voxel Grid
Methodology	Discretizes 3D point clouds into 3D voxel grid representation that preserves shape information and performs recognition using CNN or fully CNN
Limitation(s)	Empty cells in their sparse representation make it computationally inefficient, 3D convolutions result in increased inference time
Research Gap	Generating 3D region proposals could improve localization accuracy and reduce computational time