Cargando…
Discriminative Shape Feature Pooling in Deep Neural Networks
Although deep learning approaches are able to generate generic image features from massive labeled data, discriminative handcrafted features still have advantages in providing explicit domain knowledge and reflecting intuitive visual understanding. Much of the existing research focuses on integratin...
Autores principales: | , , |
---|---|
Formato: | Online Artículo Texto |
Lenguaje: | English |
Publicado: |
MDPI
2022
|
Materias: | |
Acceso en línea: | https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9144532/ https://www.ncbi.nlm.nih.gov/pubmed/35621882 http://dx.doi.org/10.3390/jimaging8050118 |
_version_ | 1784716071204290560 |
---|---|
author | Hu, Gang Dixit, Chahna Qi, Guanqiu |
author_facet | Hu, Gang Dixit, Chahna Qi, Guanqiu |
author_sort | Hu, Gang |
collection | PubMed |
description | Although deep learning approaches are able to generate generic image features from massive labeled data, discriminative handcrafted features still have advantages in providing explicit domain knowledge and reflecting intuitive visual understanding. Much of the existing research focuses on integrating both handcrafted features and deep networks to leverage the benefits. However, the issues of parameter quality have not been effectively solved in existing applications of handcrafted features in deep networks. In this research, we propose a method that enriches deep network features by utilizing the injected discriminative shape features (generic edge tokens and curve partitioning points) to adjust the network’s internal parameter update process. Thus, the modified neural networks are trained under the guidance of specific domain knowledge, and they are able to generate image representations that incorporate the benefits from both handcrafted and deep learned features. The comparative experiments were performed on several benchmark datasets. The experimental results confirmed our method works well on both large and small training datasets. Additionally, compared with existing models using either handcrafted features or deep network representations, our method not only improves the corresponding performance, but also reduces the computational costs. |
format | Online Article Text |
id | pubmed-9144532 |
institution | National Center for Biotechnology Information |
language | English |
publishDate | 2022 |
publisher | MDPI |
record_format | MEDLINE/PubMed |
spelling | pubmed-91445322022-05-29 Discriminative Shape Feature Pooling in Deep Neural Networks Hu, Gang Dixit, Chahna Qi, Guanqiu J Imaging Article Although deep learning approaches are able to generate generic image features from massive labeled data, discriminative handcrafted features still have advantages in providing explicit domain knowledge and reflecting intuitive visual understanding. Much of the existing research focuses on integrating both handcrafted features and deep networks to leverage the benefits. However, the issues of parameter quality have not been effectively solved in existing applications of handcrafted features in deep networks. In this research, we propose a method that enriches deep network features by utilizing the injected discriminative shape features (generic edge tokens and curve partitioning points) to adjust the network’s internal parameter update process. Thus, the modified neural networks are trained under the guidance of specific domain knowledge, and they are able to generate image representations that incorporate the benefits from both handcrafted and deep learned features. The comparative experiments were performed on several benchmark datasets. The experimental results confirmed our method works well on both large and small training datasets. Additionally, compared with existing models using either handcrafted features or deep network representations, our method not only improves the corresponding performance, but also reduces the computational costs. MDPI 2022-04-20 /pmc/articles/PMC9144532/ /pubmed/35621882 http://dx.doi.org/10.3390/jimaging8050118 Text en © 2022 by the authors. https://creativecommons.org/licenses/by/4.0/Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/). |
spellingShingle | Article Hu, Gang Dixit, Chahna Qi, Guanqiu Discriminative Shape Feature Pooling in Deep Neural Networks |
title | Discriminative Shape Feature Pooling in Deep Neural Networks |
title_full | Discriminative Shape Feature Pooling in Deep Neural Networks |
title_fullStr | Discriminative Shape Feature Pooling in Deep Neural Networks |
title_full_unstemmed | Discriminative Shape Feature Pooling in Deep Neural Networks |
title_short | Discriminative Shape Feature Pooling in Deep Neural Networks |
title_sort | discriminative shape feature pooling in deep neural networks |
topic | Article |
url | https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9144532/ https://www.ncbi.nlm.nih.gov/pubmed/35621882 http://dx.doi.org/10.3390/jimaging8050118 |
work_keys_str_mv | AT hugang discriminativeshapefeaturepoolingindeepneuralnetworks AT dixitchahna discriminativeshapefeaturepoolingindeepneuralnetworks AT qiguanqiu discriminativeshapefeaturepoolingindeepneuralnetworks |