A Proper Orthogonal Decomposition approach for parameters reduction of Single Shot Detector networks

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Meneghetti, Laura, Demo, Nicola, Rozza, Gianluigi
Format: Preprint
Published: 2022
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916097194721280
author Meneghetti, Laura
Demo, Nicola
Rozza, Gianluigi
author_facet Meneghetti, Laura
Demo, Nicola
Rozza, Gianluigi
contents As a major breakthrough in artificial intelligence and deep learning, Convolutional Neural Networks have achieved an impressive success in solving many problems in several fields including computer vision and image processing. Real-time performance, robustness of algorithms and fast training processes remain open problems in these contexts. In addition object recognition and detection are challenging tasks for resource-constrained embedded systems, commonly used in the industrial sector. To overcome these issues, we propose a dimensionality reduction framework based on Proper Orthogonal Decomposition, a classical model order reduction technique, in order to gain a reduction in the number of hyperparameters of the net. We have applied such framework to SSD300 architecture using PASCAL VOC dataset, demonstrating a reduction of the network dimension and a remarkable speedup in the fine-tuning of the network in a transfer learning context.
format Preprint
id arxiv_https___arxiv_org_abs_2207_13551
institution arXiv
publishDate 2022
record_format arxiv
spellingShingle A Proper Orthogonal Decomposition approach for parameters reduction of Single Shot Detector networks
Meneghetti, Laura
Demo, Nicola
Rozza, Gianluigi
Computer Vision and Pattern Recognition
Machine Learning
Numerical Analysis
As a major breakthrough in artificial intelligence and deep learning, Convolutional Neural Networks have achieved an impressive success in solving many problems in several fields including computer vision and image processing. Real-time performance, robustness of algorithms and fast training processes remain open problems in these contexts. In addition object recognition and detection are challenging tasks for resource-constrained embedded systems, commonly used in the industrial sector. To overcome these issues, we propose a dimensionality reduction framework based on Proper Orthogonal Decomposition, a classical model order reduction technique, in order to gain a reduction in the number of hyperparameters of the net. We have applied such framework to SSD300 architecture using PASCAL VOC dataset, demonstrating a reduction of the network dimension and a remarkable speedup in the fine-tuning of the network in a transfer learning context.
title A Proper Orthogonal Decomposition approach for parameters reduction of Single Shot Detector networks
topic Computer Vision and Pattern Recognition
Machine Learning
Numerical Analysis
url https://arxiv.org/abs/2207.13551