|
Title:
|
HYPER-PARAMETER TUNING WITH AUTOMATED MACHINE LEARNING FOR POINT CLOUD PART SEGMENTATION |
|
Author(s):
|
Gabriel Lenz Balatka and Rafael Stubs Parpinelli |
|
ISBN:
|
978-989-8704-71-9 |
|
Editors:
|
Paula Miranda and Pedro Isaías |
|
Year:
|
2025 |
|
Edition:
|
Single |
|
Keywords:
|
Point Cloud Subsampling, Point Cloud Part Segmentation, Automated Machine Learning, Artificial Neural Networks,
Hyper-Parameter Tuning |
|
Type:
|
Full Paper |
|
First Page:
|
83 |
|
Last Page:
|
90 |
|
Language:
|
English |
|
Cover:
|
|
|
Full Contents:
|
if you are a member please login 
|
|
Paper Abstract:
|
This study aimed to investigate the contributions of hyper-parameter optimization methods for Artificial Neural Networks
applied to the Point Cloud Part Segmentation process. The PartNet database was used, selecting three specific categories,
Chair-1, StorageFurniture-1 and Lamp-1, through data analysis of each category. For each of them, the Furthest Point
Sampling algorithm (FPS) was used, creating two data sets, with each point cloud containing 256 and 512 points. The
neural network architecture selected was PointNet, which has historically shown promise for the task. The AutoML strategy
was used for hyper-parameter adjustment. As a result, it was possible to see the importance of hyper-parameter tuning and
also discuss the impact of choosing the number of points in each point cloud. |
|
|
|
|
|
|
