Stop-band Energy Constraint for Orthogonal Tunable Wavelet Units in Convolutional Neural Networks for Computer Vision problems

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Le, An D., Nguyen, Hung, Seo, Sungbal, Bae, You-Suk, Nguyen, Truong Q.
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911069450010624
author Le, An D.
Nguyen, Hung
Seo, Sungbal
Bae, You-Suk
Nguyen, Truong Q.
author_facet Le, An D.
Nguyen, Hung
Seo, Sungbal
Bae, You-Suk
Nguyen, Truong Q.
contents This work introduces a stop-band energy constraint for filters in orthogonal tunable wavelet units with a lattice structure, aimed at improving image classification and anomaly detection in CNNs, especially on texture-rich datasets. Integrated into ResNet-18, the method enhances convolution, pooling, and downsampling operations, yielding accuracy gains of 2.48% on CIFAR-10 and 13.56% on the Describable Textures dataset. Similar improvements are observed in ResNet-34. On the MVTec hazelnut anomaly detection task, the proposed method achieves competitive results in both segmentation and detection, outperforming existing approaches.
format Preprint
id arxiv_https___arxiv_org_abs_2507_16114
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Stop-band Energy Constraint for Orthogonal Tunable Wavelet Units in Convolutional Neural Networks for Computer Vision problems
Le, An D.
Nguyen, Hung
Seo, Sungbal
Bae, You-Suk
Nguyen, Truong Q.
Computer Vision and Pattern Recognition
Signal Processing
This work introduces a stop-band energy constraint for filters in orthogonal tunable wavelet units with a lattice structure, aimed at improving image classification and anomaly detection in CNNs, especially on texture-rich datasets. Integrated into ResNet-18, the method enhances convolution, pooling, and downsampling operations, yielding accuracy gains of 2.48% on CIFAR-10 and 13.56% on the Describable Textures dataset. Similar improvements are observed in ResNet-34. On the MVTec hazelnut anomaly detection task, the proposed method achieves competitive results in both segmentation and detection, outperforming existing approaches.
title Stop-band Energy Constraint for Orthogonal Tunable Wavelet Units in Convolutional Neural Networks for Computer Vision problems
topic Computer Vision and Pattern Recognition
Signal Processing
url https://arxiv.org/abs/2507.16114