Masked Intra-class Consistency Feature Reconstruction Network for Few-Shot Fine-Grained Classification.

Saved in:

Bibliographic Details
Title:	Masked Intra-class Consistency Feature Reconstruction Network for Few-Shot Fine-Grained Classification.
Authors:	Wu, Jijie¹ jijie@lut.edu.cn, Yin, Qiyu¹ yqy@lut.edu.cn, Li, Xiaoxu¹ lixiaoxu@lut.edu.cn
Source:	IAENG International Journal of Computer Science. May2026, Vol. 53 Issue 5, p1666-1676. 11p.
Subjects:	Attention, Image recognition (Computer vision), Machine learning
Abstract:	Fine-grained few-shot image classification requires recognizing subtle inter-class differences with only a limited number of samples. Reconstruction-based methods improve metric accuracy through spatial feature alignment, effectively reducing intra-class variation and enhancing classification performance. However, existing methods indiscriminately align all features, which introduces task-irrelevant noise and leads to "over-alignment," while also failing to leverage self-supervisory signals to enhance intra-class consistency fully. To address these issues, we propose the Masked Intra-class Consistency Feature Reconstruction Network (MIC-FRN). Specifically, we introduce a Mask-Guided Attention Module (MGAM) into the reconstruction framework, where random masking combined with learnable attention suppresses irrelevant regions and serves as regularization. In addition, we further design an Intra-Class Consistency Module (ICCM) using a self-alignment mechanism for support samples under different masked views, enabling the model to maintain intra-class consistency under varying feature visibility. Experimental results on three benchmark fine-grained datasets show that our method significantly outperforms existing reconstruction networks and other baselines in both 5-way 1-shot and 5-shot tasks, validating the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]
	Copyright of IAENG International Journal of Computer Science is the property of International Association of Engineers (IAENG) and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database:	Engineering Source

Description
Abstract:	Fine-grained few-shot image classification requires recognizing subtle inter-class differences with only a limited number of samples. Reconstruction-based methods improve metric accuracy through spatial feature alignment, effectively reducing intra-class variation and enhancing classification performance. However, existing methods indiscriminately align all features, which introduces task-irrelevant noise and leads to "over-alignment," while also failing to leverage self-supervisory signals to enhance intra-class consistency fully. To address these issues, we propose the Masked Intra-class Consistency Feature Reconstruction Network (MIC-FRN). Specifically, we introduce a Mask-Guided Attention Module (MGAM) into the reconstruction framework, where random masking combined with learnable attention suppresses irrelevant regions and serves as regularization. In addition, we further design an Intra-Class Consistency Module (ICCM) using a self-alignment mechanism for support samples under different masked views, enabling the model to maintain intra-class consistency under varying feature visibility. Experimental results on three benchmark fine-grained datasets show that our method significantly outperforms existing reconstruction networks and other baselines in both 5-way 1-shot and 5-shot tasks, validating the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]
ISSN:	1819656X