Hybrid Multi-Granularity Approach for Few-Shot Image Retrieval with Weak Features

This paper proposes a multi-granularity retrieval algorithm based on an unsupervised image augmentation network. The algorithm designs a feature extraction method (AugODNet_BRA) rooted in image augmentation, which efficiently captures high-level semantic features of images with few samples, small ta...

Full description

Saved in:
Bibliographic Details
Main Authors: Aiguo Lu, Zican Li, Yanwei Liu, Pandi Liu, Ke Wang
Format: Article
Language:English
Published: MDPI AG 2025-05-01
Series:Algorithms
Subjects:
Online Access:https://www.mdpi.com/1999-4893/18/6/329
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850157064131706880
author Aiguo Lu
Zican Li
Yanwei Liu
Pandi Liu
Ke Wang
author_facet Aiguo Lu
Zican Li
Yanwei Liu
Pandi Liu
Ke Wang
author_sort Aiguo Lu
collection DOAJ
description This paper proposes a multi-granularity retrieval algorithm based on an unsupervised image augmentation network. The algorithm designs a feature extraction method (AugODNet_BRA) rooted in image augmentation, which efficiently captures high-level semantic features of images with few samples, small targets, and weak features through unsupervised learning. The Omni-Dimensional Dynamic Convolution module and Bi-Level Routing Attention mechanism are introduced to enhance the model’s adaptability to complex scenes and variable features, thereby improving its capability to capture details of small targets. The Omni-Dimensional Dynamic Convolution module flexibly adjusts the dimensions of convolution kernels to accommodate small targets of varying sizes and shapes. At the same time, the Bi-Level Routing Attention mechanism adaptively focuses on key regions, boosting the model’s discriminative ability for targets in complex backgrounds. The optimized loss function further enhances the robustness and distinctiveness of features, improving retrieval accuracy. The experimental results demonstrate that the proposed method outperforms baseline algorithms on the public dataset CUB-200-2011 and exhibits great potential for application and practical value in scenarios such as carrier-based aircraft tail hook recognition.
format Article
id doaj-art-cd1491a4fafd4169ac71cb6f9fd350ae
institution OA Journals
issn 1999-4893
language English
publishDate 2025-05-01
publisher MDPI AG
record_format Article
series Algorithms
spelling doaj-art-cd1491a4fafd4169ac71cb6f9fd350ae2025-08-20T02:24:17ZengMDPI AGAlgorithms1999-48932025-05-0118632910.3390/a18060329Hybrid Multi-Granularity Approach for Few-Shot Image Retrieval with Weak FeaturesAiguo Lu0Zican Li1Yanwei Liu2Pandi Liu3Ke Wang4School of Mechanical Engineering, Southeast University, Nanjing 211189, ChinaSchool of Computer and Artificial Intelligence, Zhengzhou University, Zhengzhou 450001, ChinaSchool of Computer and Artificial Intelligence, Zhengzhou University, Zhengzhou 450001, ChinaSchool of Computer and Artificial Intelligence, Zhengzhou University, Zhengzhou 450001, ChinaSchool of Computer and Artificial Intelligence, Zhengzhou University, Zhengzhou 450001, ChinaThis paper proposes a multi-granularity retrieval algorithm based on an unsupervised image augmentation network. The algorithm designs a feature extraction method (AugODNet_BRA) rooted in image augmentation, which efficiently captures high-level semantic features of images with few samples, small targets, and weak features through unsupervised learning. The Omni-Dimensional Dynamic Convolution module and Bi-Level Routing Attention mechanism are introduced to enhance the model’s adaptability to complex scenes and variable features, thereby improving its capability to capture details of small targets. The Omni-Dimensional Dynamic Convolution module flexibly adjusts the dimensions of convolution kernels to accommodate small targets of varying sizes and shapes. At the same time, the Bi-Level Routing Attention mechanism adaptively focuses on key regions, boosting the model’s discriminative ability for targets in complex backgrounds. The optimized loss function further enhances the robustness and distinctiveness of features, improving retrieval accuracy. The experimental results demonstrate that the proposed method outperforms baseline algorithms on the public dataset CUB-200-2011 and exhibits great potential for application and practical value in scenarios such as carrier-based aircraft tail hook recognition.https://www.mdpi.com/1999-4893/18/6/329multi-granularityunsupervised image augmentation networkOmni-Dimensional Dynamic ConvolutionBi-Level Routing Attention
spellingShingle Aiguo Lu
Zican Li
Yanwei Liu
Pandi Liu
Ke Wang
Hybrid Multi-Granularity Approach for Few-Shot Image Retrieval with Weak Features
Algorithms
multi-granularity
unsupervised image augmentation network
Omni-Dimensional Dynamic Convolution
Bi-Level Routing Attention
title Hybrid Multi-Granularity Approach for Few-Shot Image Retrieval with Weak Features
title_full Hybrid Multi-Granularity Approach for Few-Shot Image Retrieval with Weak Features
title_fullStr Hybrid Multi-Granularity Approach for Few-Shot Image Retrieval with Weak Features
title_full_unstemmed Hybrid Multi-Granularity Approach for Few-Shot Image Retrieval with Weak Features
title_short Hybrid Multi-Granularity Approach for Few-Shot Image Retrieval with Weak Features
title_sort hybrid multi granularity approach for few shot image retrieval with weak features
topic multi-granularity
unsupervised image augmentation network
Omni-Dimensional Dynamic Convolution
Bi-Level Routing Attention
url https://www.mdpi.com/1999-4893/18/6/329
work_keys_str_mv AT aiguolu hybridmultigranularityapproachforfewshotimageretrievalwithweakfeatures
AT zicanli hybridmultigranularityapproachforfewshotimageretrievalwithweakfeatures
AT yanweiliu hybridmultigranularityapproachforfewshotimageretrievalwithweakfeatures
AT pandiliu hybridmultigranularityapproachforfewshotimageretrievalwithweakfeatures
AT kewang hybridmultigranularityapproachforfewshotimageretrievalwithweakfeatures