Huayu Fu , Guosheng Xin , Yanshen Zhao , Cong Wang , Zhongzhi Han
{"title":"Efficient sweet potato quality detection using pruned attention-based deep learning","authors":"Huayu Fu , Guosheng Xin , Yanshen Zhao , Cong Wang , Zhongzhi Han","doi":"10.1016/j.jspr.2025.102749","DOIUrl":null,"url":null,"abstract":"<div><div>Sweet potatoes are rich in essential nutrients crucial for maintaining human health, and their market value is primarily determined by external quality. To address the classification of sweet potato appearance quality, we propose a rapid assessment method for evaluating external quality. Using a dataset of 1833 images of the <em>Yanshu 25</em> variety, we adopt the MobileNetV4 series network with a reduced Universal Inverted Bottleneck (UIB) module as the core architecture. The network is trained and pruned by integrating a learning rate scheduler, a multi-head self-attention (MHSA) mechanism, and a pruning algorithm based on gated mechanisms. After pruning, the optimal model achieves an accuracy of 93.08 % across five categories, with an average detection speed of only 0.0022 s per sample. This method offers outstanding detection accuracy, faster processing speed, and significantly reduced model complexity, making it highly effective for sweet potato quality assessment.</div></div>","PeriodicalId":17019,"journal":{"name":"Journal of Stored Products Research","volume":"114 ","pages":"Article 102749"},"PeriodicalIF":2.7000,"publicationDate":"2025-07-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of Stored Products Research","FirstCategoryId":"97","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0022474X25002085","RegionNum":2,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"ENTOMOLOGY","Score":null,"Total":0}
引用次数: 0
Abstract
Sweet potatoes are rich in essential nutrients crucial for maintaining human health, and their market value is primarily determined by external quality. To address the classification of sweet potato appearance quality, we propose a rapid assessment method for evaluating external quality. Using a dataset of 1833 images of the Yanshu 25 variety, we adopt the MobileNetV4 series network with a reduced Universal Inverted Bottleneck (UIB) module as the core architecture. The network is trained and pruned by integrating a learning rate scheduler, a multi-head self-attention (MHSA) mechanism, and a pruning algorithm based on gated mechanisms. After pruning, the optimal model achieves an accuracy of 93.08 % across five categories, with an average detection speed of only 0.0022 s per sample. This method offers outstanding detection accuracy, faster processing speed, and significantly reduced model complexity, making it highly effective for sweet potato quality assessment.
期刊介绍:
The Journal of Stored Products Research provides an international medium for the publication of both reviews and original results from laboratory and field studies on the preservation and safety of stored products, notably food stocks, covering storage-related problems from the producer through the supply chain to the consumer. Stored products are characterised by having relatively low moisture content and include raw and semi-processed foods, animal feedstuffs, and a range of other durable items, including materials such as clothing or museum artefacts.