Extracting meaning from 3D data is not a simple objective. We start with massive datasets rich in geometric details from which we want to extract a semantic understanding (Chapter 1) needed for high-level tasks. The first challenge is the disparity between raw data and usable knowledge that constrains the potential of 3D machine learning across multiple disciplines, including robotics, autonomous navigation, and scene comprehension.
Indeed, while we benefit from an immense accessible corpus for training supervised models for image or text modalities, 3D openly accessible labeled datasets are rare. The lack of inherent labels makes supervised learning methods challenging or too expensive to implement on a large scale.
We require a method to bridge this gap without depending extensively on costly and labor-intensive human labeling. This feeds the idea to enable machines to autonomously discern the intrinsic structure within 3D data. This would enhance efficiency and allow us to address numerous difficulties. I want to teach you how to achieve this using the power of unsupervised learning through clustering algorithms (see Figure 12-1).
In this chapter, we focus on clustering through unsupervised learning mechanisms. Chapters 6 and 7 provide dimensionality reduction solutions; Chapters 8 and 9 showcase the use of regression techniques; and Chapters 14, 15, and16 are a deep dive into classification approaches.
Figure 12-1. The main differences ...
Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month, and much more.
