A survey on Deep Learning Advances on Different 3D Data Representations

TL;DR

This survey reviews recent advances in applying deep learning to various 3D data representations, discussing challenges and differences between Euclidean and non-Euclidean formats in tasks like segmentation and recognition.

Contribution

It provides a comprehensive overview of 3D data representations and analyzes how deep learning techniques are adapted for each, highlighting key challenges and differences.

Findings

Deep learning has been successfully applied to 3D data tasks.

Different 3D data representations pose unique challenges for deep learning.

Euclidean and non-Euclidean data require distinct approaches.

Abstract

3D data is a valuable asset the computer vision filed as it provides rich information about the full geometry of sensed objects and scenes. Recently, with the availability of both large 3D datasets and computational power, it is today possible to consider applying deep learning to learn specific tasks on 3D data such as segmentation, recognition and correspondence. Depending on the considered 3D data representation, different challenges may be foreseen in using existent deep learning architectures. In this work, we provide a comprehensive overview about various 3D data representations highlighting the difference between Euclidean and non-Euclidean ones. We also discuss how Deep Learning methods are applied on each representation, analyzing the challenges to overcome.