Accurate and real-time fish feeding intensity scoring using channel convolution GLU and SMFA attention fusion in recirculating aquaculture system

IF 3.9 1区农林科学 Q1 FISHERIES

Aquaculture Pub Date : 2025-09-10 DOI:10.1016/j.aquaculture.2025.743180

Chenjian Liu , Xinting Yang , Baoliang Liu , Zhenxi Zhao , Pingchuan Ma , Tingting Fu , Weichen Hu , Xiaoqiang Gao , Chao Zhou

{"title":"Accurate and real-time fish feeding intensity scoring using channel convolution GLU and SMFA attention fusion in recirculating aquaculture system","authors":"Chenjian Liu , Xinting Yang , Baoliang Liu , Zhenxi Zhao , Pingchuan Ma , Tingting Fu , Weichen Hu , Xiaoqiang Gao , Chao Zhou","doi":"10.1016/j.aquaculture.2025.743180","DOIUrl":null,"url":null,"abstract":"<div><div>In aquaculture, real-time quantification of fish feeding intensity is critical for developing scientific feeding strategies. Previous deep learning-based studies primarily focused on rough recognition of feeding and non-feeding patterns. However, with the growing demand for intelligent feeding systems, there is an urgent need for precise and quantitative assessment of fish feeding intensity. To address this, this study proposes CS-TransNeXt, a fish feeding intensity scoring model that integrates Channel Convolution GLU (CCGLU) and Self-Modulation Feature Aggregation (SMFA), which can precisely quantify feeding intensity into 10 scoring scales (1−10). Specifically, the CCGLU module is introduced into the TransNeXt, so as to enhance local feature modeling by fusing multi-scale Depthwise Convolutions with channel attention. Meanwhile, the SMFA replaces the original multi-head self-attention in TransNeXt, enabling adaptive weight adjustment through variance-based dynamic parameters of global-local features. Experimental results demonstrate that the proposed CS-TransNeXt achieves a Top-1 Accuracy of 95.25 %, an F1-Score of 95.30 %, outperforming the baseline TransNeXt-micro by 4.00 % in accuracy. Meanwhile, it is only 17.60 M, and provides a novel method for high-precision quantitative scoring of fish feeding intensity, offering significant practical value for the development of intelligent feeding systems.</div></div>","PeriodicalId":8375,"journal":{"name":"Aquaculture","volume":"612 ","pages":"Article 743180"},"PeriodicalIF":3.9000,"publicationDate":"2025-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Aquaculture","FirstCategoryId":"97","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S004484862501066X","RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"FISHERIES","Score":null,"Total":0}

引用次数: 0

Abstract

In aquaculture, real-time quantification of fish feeding intensity is critical for developing scientific feeding strategies. Previous deep learning-based studies primarily focused on rough recognition of feeding and non-feeding patterns. However, with the growing demand for intelligent feeding systems, there is an urgent need for precise and quantitative assessment of fish feeding intensity. To address this, this study proposes CS-TransNeXt, a fish feeding intensity scoring model that integrates Channel Convolution GLU (CCGLU) and Self-Modulation Feature Aggregation (SMFA), which can precisely quantify feeding intensity into 10 scoring scales (1−10). Specifically, the CCGLU module is introduced into the TransNeXt, so as to enhance local feature modeling by fusing multi-scale Depthwise Convolutions with channel attention. Meanwhile, the SMFA replaces the original multi-head self-attention in TransNeXt, enabling adaptive weight adjustment through variance-based dynamic parameters of global-local features. Experimental results demonstrate that the proposed CS-TransNeXt achieves a Top-1 Accuracy of 95.25 %, an F1-Score of 95.30 %, outperforming the baseline TransNeXt-micro by 4.00 % in accuracy. Meanwhile, it is only 17.60 M, and provides a novel method for high-precision quantitative scoring of fish feeding intensity, offering significant practical value for the development of intelligent feeding systems.

查看原文本刊更多论文

基于通道卷积GLU和SMFA注意力融合的循循水养殖系统鱼类摄食强度准确实时评分

在水产养殖中，鱼类摄食强度的实时量化对于制定科学的摄食策略至关重要。以前基于深度学习的研究主要集中在对喂食和非喂食模式的粗略识别上。然而，随着人们对智能饲养系统的需求日益增长，迫切需要对鱼类的饲养强度进行精确和定量的评估。为了解决这个问题，本研究提出了CS-TransNeXt，这是一个整合了通道卷积GLU （CCGLU）和自调制特征聚合（SMFA）的鱼类进食强度评分模型，可以精确地将进食强度量化为10个评分尺度（1−10）。具体而言，在TransNeXt中引入CCGLU模块，通过融合多尺度深度卷积和信道关注来增强局部特征建模。同时，SMFA取代了TransNeXt中原有的多头自关注，通过基于方差的全局-局部特征动态参数实现自适应权值调整。实验结果表明，本文提出的CS-TransNeXt的Top-1准确率为95.25%，F1-Score为95.30%，比基线TransNeXt-micro的准确率提高了4.00%。同时仅为17.60 M，为鱼类摄食强度的高精度定量评分提供了一种新颖的方法，对智能摄食系统的开发具有重要的实用价值。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

Aquaculture 农林科学-海洋与淡水生物学

CiteScore

8.60

自引率

17.80%

发文量

1246

审稿时长

56 days

期刊介绍： Aquaculture is an international journal for the exploration, improvement and management of all freshwater and marine food resources. It publishes novel and innovative research of world-wide interest on farming of aquatic organisms, which includes finfish, mollusks, crustaceans and aquatic plants for human consumption. Research on ornamentals is not a focus of the Journal. Aquaculture only publishes papers with a clear relevance to improving aquaculture practices or a potential application.