PRS2Net：基于过滤器修剪和注意机制的高效智能胡萝卜检测模型。

IF 3.5 2区农林科学 Q1 AGRICULTURE, MULTIDISCIPLINARY

Journal of the Science of Food and Agriculture Pub Date : 2025-08-22 DOI:10.1002/jsfa.70142

Huayu Fu, Hongfei Zhu, Yifan Zhao, Hang Liu, Xuetong Zhai, Cong Wang, Yanshen Zhao, Limiao Deng, Zhongzhi Han

{"title":"PRS2Net：基于过滤器修剪和注意机制的高效智能胡萝卜检测模型。","authors":"Huayu Fu, Hongfei Zhu, Yifan Zhao, Hang Liu, Xuetong Zhai, Cong Wang, Yanshen Zhao, Limiao Deng, Zhongzhi Han","doi":"10.1002/jsfa.70142","DOIUrl":null,"url":null,"abstract":"Background: Carrots, rich in essential nutrients, play a crucial role in supporting human health. Current deep learning networks for carrot quality inspection are constrained by redundant parameters and high computational costs. To address these issues, this paper introduces a lightweight network, PRS2Net, based on ResNet18 selected after comparing four networks (GoogLeNet, MobileNet-v2, ResNet18, ResNet50). ResNet18 was pruned using first-order Taylor expansion to reduce redundancy and enhanced with an attention mechanism to focus on critical features.Results: PRS2Net achieved high efficiency with learnable parameters reduced from 11 173 764 to 444 152, while maintaining 97.25% accuracy on the validation set. Training time was cut by about 53.15% compared to ResNet18, significantly speeding up carrot quality inspection.Conclusion: This approach enhances the speed and efficiency of carrot quality evaluation, offering a practical, resource-efficient solution for real-world applications in agriculture and food industries, potentially reducing operational costs and improving scalability for automated systems. © 2025 Society of Chemical Industry.","PeriodicalId":17725,"journal":{"name":"Journal of the Science of Food and Agriculture","volume":" ","pages":""},"PeriodicalIF":3.5000,"publicationDate":"2025-08-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"PRS2Net: an efficient intelligent carrot detection model via filter pruning and attention mechanisms.\",\"authors\":\"Huayu Fu, Hongfei Zhu, Yifan Zhao, Hang Liu, Xuetong Zhai, Cong Wang, Yanshen Zhao, Limiao Deng, Zhongzhi Han\",\"doi\":\"10.1002/jsfa.70142\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Background: Carrots, rich in essential nutrients, play a crucial role in supporting human health. Current deep learning networks for carrot quality inspection are constrained by redundant parameters and high computational costs. To address these issues, this paper introduces a lightweight network, PRS2Net, based on ResNet18 selected after comparing four networks (GoogLeNet, MobileNet-v2, ResNet18, ResNet50). ResNet18 was pruned using first-order Taylor expansion to reduce redundancy and enhanced with an attention mechanism to focus on critical features.Results: PRS2Net achieved high efficiency with learnable parameters reduced from 11 173 764 to 444 152, while maintaining 97.25% accuracy on the validation set. Training time was cut by about 53.15% compared to ResNet18, significantly speeding up carrot quality inspection.Conclusion: This approach enhances the speed and efficiency of carrot quality evaluation, offering a practical, resource-efficient solution for real-world applications in agriculture and food industries, potentially reducing operational costs and improving scalability for automated systems. © 2025 Society of Chemical Industry.\",\"PeriodicalId\":17725,\"journal\":{\"name\":\"Journal of the Science of Food and Agriculture\",\"volume\":\" \",\"pages\":\"\"},\"PeriodicalIF\":3.5000,\"publicationDate\":\"2025-08-22\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Journal of the Science of Food and Agriculture\",\"FirstCategoryId\":\"97\",\"ListUrlMain\":\"https://doi.org/10.1002/jsfa.70142\",\"RegionNum\":2,\"RegionCategory\":\"农林科学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q1\",\"JCRName\":\"AGRICULTURE, MULTIDISCIPLINARY\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of the Science of Food and Agriculture","FirstCategoryId":"97","ListUrlMain":"https://doi.org/10.1002/jsfa.70142","RegionNum":2,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"AGRICULTURE, MULTIDISCIPLINARY","Score":null,"Total":0}

引用次数: 0

摘要

背景：胡萝卜富含人体必需营养素，对维持人体健康起着至关重要的作用。目前用于胡萝卜质量检测的深度学习网络受冗余参数和高计算成本的限制。为了解决这些问题，本文介绍了一种基于ResNet18的轻量级网络PRS2Net，该网络是在比较了四种网络（GoogLeNet, MobileNet-v2, ResNet18, ResNet50）后选择的。使用一阶Taylor展开对ResNet18进行修剪以减少冗余，并使用关注机制对其进行增强，以关注关键特征。结果：PRS2Net获得了较高的效率，可学习参数从11 173 764减少到444 152，同时在验证集上保持了97.25%的准确率。与ResNet18相比，培训时间缩短约53.15%，显著加快了胡萝卜质量检测速度。结论：该方法提高了胡萝卜质量评价的速度和效率，为农业和食品工业的实际应用提供了一种实用的、资源高效的解决方案，有可能降低操作成本，提高自动化系统的可扩展性。©2025化学工业协会。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

查看原文本刊更多论文

PRS2Net: an efficient intelligent carrot detection model via filter pruning and attention mechanisms.

Background: Carrots, rich in essential nutrients, play a crucial role in supporting human health. Current deep learning networks for carrot quality inspection are constrained by redundant parameters and high computational costs. To address these issues, this paper introduces a lightweight network, PRS2Net, based on ResNet18 selected after comparing four networks (GoogLeNet, MobileNet-v2, ResNet18, ResNet50). ResNet18 was pruned using first-order Taylor expansion to reduce redundancy and enhanced with an attention mechanism to focus on critical features.

Results: PRS2Net achieved high efficiency with learnable parameters reduced from 11 173 764 to 444 152, while maintaining 97.25% accuracy on the validation set. Training time was cut by about 53.15% compared to ResNet18, significantly speeding up carrot quality inspection.

Conclusion: This approach enhances the speed and efficiency of carrot quality evaluation, offering a practical, resource-efficient solution for real-world applications in agriculture and food industries, potentially reducing operational costs and improving scalability for automated systems. © 2025 Society of Chemical Industry.

求助全文

通过发布文献求助，成功后即可免费获取论文全文。去求助

来源期刊

Journal of the Science of Food and Agriculture 工程技术-农业综合

CiteScore

8.10

自引率

4.90%

发文量

634

审稿时长

3.1 months

期刊介绍： The Journal of the Science of Food and Agriculture publishes peer-reviewed original research, reviews, mini-reviews, perspectives and spotlights in these areas, with particular emphasis on interdisciplinary studies at the agriculture/ food interface. Published for SCI by John Wiley & Sons Ltd. SCI (Society of Chemical Industry) is a unique international forum where science meets business on independent, impartial ground. Anyone can join and current Members include consumers, business people, environmentalists, industrialists, farmers, and researchers. The Society offers a chance to share information between sectors as diverse as food and agriculture, pharmaceuticals, biotechnology, materials, chemicals, environmental science and safety. As well as organising educational events, SCI awards a number of prestigious honours and scholarships each year, publishes peer-reviewed journals, and provides Members with news from their sectors in the respected magazine, Chemistry & Industry . Originally established in London in 1881 and in New York in 1894, SCI is a registered charity with Members in over 70 countries.