RFHNet Relational and Frequency-Aware Hashing Network for Large-Scale Fine-Grained Food Image Retrieval

Jul 16, 2026

Upload by Rowan

Fine-grained food image retrieval is essential in computational gastronomy, enabling food traceability, dietary monitoring, and smart catering. Although hashing methods offer storage efficiency and fast Hamming-distance search, they often underperform when subtle local semantics and frequency-sensitive cues determine similarity. RFHNet introduces a cascaded hierarchical hashing network with Fine-grained Relation Modeling, Multi-Frequency Modulated Fusion, and Hierarchical Semantic Synergy to integrate global and local signals into discriminative codes. Experiments on six benchmarks show mAP gains of 4.44%–17.20% at 12 bits, outperforming state-of-the-art.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages10

Reading Time25 min

Introduction

Problem background and challenges

Fine-grained image retrieval and hashing approaches

Limitations of existing methods

Method

Fine-grained Relation Modeling (FRM)

Multi-Frequency Modulated Fusion (MFMF)

Hierarchical Semantic Synergy (HSS)

Experiments

Benchmark results and comparisons

Conclusion

RFHNet: Relational and Frequency-Aware Hashing Network for Large-Scale Fine-Grained Food Image Retrieval Junsong Wang College of Computer Science and Artificial Intelligence, Ludong University Yantai, China [wangjunsong@m.ldu.edu.cn](wangjunsong@m.ldu.edu.cn) Weiqing Min Institute of Computing Technology, Chinese Academy of Sciences Beijing, China [minweiqing@ict.ac.cn](minweiqing@ict.ac.cn) Guorui Sheng College of Computer Science and Artificial Intelligence, Ludong University Yantai, China [shengguorui@ldu.edu.cn](shengguorui@ldu.edu.cn) Tao Yao College of Computer Science and Artificial Intelligence, Ludong University Yantai, China [yaotao@ldu.edu.cn](yaotao@ldu.edu.cn) Lili Wang∗ College of Computer Science and Artificial Intelligence, Ludong University Yantai, China [wanglili@ldu.edu.cn](wanglili@ldu.edu.cn) Shuqiang Jiang University of Chinese Academy of Sciences Beijing, China [sqjiang@ict.ac.cn](sqjiang@ict.ac.cn) arXiv :2607 .06 148v 1 [ cs .CV] 7 Jul 2026 Abstract Fine-grained food image retrieval is a key task in computational gastronomy, with applications in food traceability, dietary monitoring, and smart catering systems. Although hashing-based retrieval is attractive for large-scale search due to its storage efficiency and fast Hamming-distance computation, existing methods often perform poorly in fine-grained food scenarios, where subtle local semantics and frequency-sensitive visual cues are essential. To address this challenge, we propose RFHNet, a cascaded hierarchical hashing network that captures both global structure and fine-grained local details through multi-level representations. RFHNet includes three components: (1) Fine-grained Relation Modeling (FRM) to capture subtle visual differences among similar food components; (2) Multi-Frequency Modulated Fusion (MFMF) to extract informative multi-frequency features; and (3) Hierarchical Semantic Synergy (HSS) to adaptively integrate multi-level representations and generate discriminative hash codes. Experiments on six food-specific benchmarks show that RFHNet consistently outperforms state-ofthe-art hashing methods, with mAP gains of 4.44% to 17.20% at 12 bits. These results validate the effectiveness of RFHNet for largescale visual food retrieval and smart catering applications. The source code will be released upon publication. CCS Concepts • Information systems → Image search; • Computing methodologies → Visual content-based indexing and retrieval. Keywords Fine-grained image retrieval, hash learning, multi-frequency information, food computing. ∗ Corresponding author This work is licensed under a Creative Commons Attribution 4 .0 International License. ICMR’26, Amsterdam, Netherlands © 2026 Copyright held by the owner/author(s) . ACM ISBN 979-8-4007-2617-0/2026/06 [https://doi.org/10.1145/3805622.3810838](https://doi.org/10.1145/3805622.3810838) ACM Reference Format: Junsong Wang, Weiqing Min, Guorui Sheng, Tao Yao, Lili Wang, and Shuqiang Jiang. 2026. RFHNet: Relational and Frequency-Aware Hashing Network for Large-Scale Fine-Grained Food Image Retrieval. In International Conference on Multimedia Retrieval (ICMR’26), June 16–19, 2026, Amsterdam, Netherlands. ACM, New York, NY, USA, 10 pages. [https://doi.org/10.1145/](https://doi.org/10.1145/)[ ](https://doi.org/10.1145/)3805622.3810838 1 Introduction Food image retrieval is a fundamental task in food computing, supporting applications such as intelligent menu recommendation, dietary monitoring, and nutrition management [16, 32] . Compared with generic image retrieval, food-related scenarios pose unique challenges due to fine-grained visual variations caused by ingredient composition, cooking styles, and presentation conditions. These challenges are further amplified in large-scale settings, where efficient and discriminative retrieval becomes critical. With the increasing scale of food image datasets, Fine-Grained Image Retrieval (FGIR) [48] has

FAQs

What problem does RFHNet target?

RFHNet targets large-scale fine-grained food image retrieval, where visually similar dishes and subtle local variations make discrimination difficult.

Why do existing hashing methods struggle in fine-grained food scenarios?

Existing methods often fail to capture subtle spatial relations and are sensitive to both high-frequency noise and low-frequency artifacts, leading to unstable representations.

What are the main components of RFHNet?

RFHNet consists of Fine-grained Relation Modeling (FRM), Multi-Frequency Modulated Fusion (MFMF), and Hierarchical Semantic Synergy (HSS) to model relations, extract multi-frequency features, and integrate multi-level representations for discriminative hash codes.

Jul 16, 2026

Upload by Rowan