Fish‐Finder: A robust small target detection method for aquaculture fish in low‐quality underwater images DOI
Liang Liu, Junfeng Wu,

Haiyan Zhao

и другие.

Journal of Fish Biology, Год журнала: 2024, Номер unknown

Опубликована: Ноя. 25, 2024

Abstract Underwater fish object detection serves as a pivotal research direction in marine biology, aquaculture management, and computer vision, yet it poses substantial challenges due to the complexity of underwater environments, occultations, small‐sized frequently moving aquaculture. Addressing these challenges, we propose novel algorithm named Fish‐Finder. First, engendered structure titled “C2fBF,” utilizing dual‐path routing attention protocol BiFormer. The primary objective this is alleviate perturbations induced by intricacies during phase downsampling backbone network, thereby discerning conserving finer contextual features. Subsequently, co‐opted RepGFPN method within our neck network—a distinctive approach that adeptly merges high‐level semantic constructs with low‐level spatial specifics, thus fortifying its multi‐scale prowess. Then, an endeavor diminish sensitivity toward positional aberrations diminutive aquatic creatures, incorporated bounding box regression loss function, Wasserstein loss, existing CIoU. This innovative function gauges congruity between predicted Gaussian distribution reference distribution. Finally, regard dataset, independently assembled specific dataset termed “SmallFish.” unique meticulously designed for small‐scale intricate settings, includes 5000 annotated images small fish. Experimental results demonstrate that, compared state‐of‐the‐art methods, proposed improves accuracy , mean average precision (mAP) increases public Kaggle‐Fish SmallFish respectively.

Язык: Английский

Hybridizing Long Short-Term Memory and Bi-Directional Long Short-Term Memory Models for Efficient Classification: A Study on Xanthomonas axonopodis pv. phaseoli (XaP) in Two Bean Varieties DOI Creative Commons
Ramazan Kursun, A. Gur, Kubilay Kurtuluş Bastas

и другие.

Agronomy, Год журнала: 2024, Номер 14(7), С. 1495 - 1495

Опубликована: Июль 10, 2024

This study was conducted on Xanthomonas axonopodis pv, which causes significant economic losses in the agricultural sector. Here, we a common bacterial blight disease caused by phaseoli (XaP) pathogen Üstün42 and Akbulut bean genera. In this study, total of 4000 images, healthy diseased, were used for both breeds. These images classified AlexNet, VGG16, VGG19 models. Later, reclassification performed applying pre-processing to raw images. According results obtained, accuracy rates pre-processed VGG19, VGG16 AlexNet models determined as 0.9213, 0.9125 0.8950, respectively. The then hybridized with LSTM BiLSTM new created. When performance these hybrid evaluated, it found that more successful than simple models, while gave better LSTM. particular, VGG19+BiLSTM model attracted attention achieving 94.25% classification emphasizes effectiveness image processing techniques agriculture field detection is important dataset literature evaluating

Язык: Английский

Процитировано

0
Automated Visual Quality Detection for Tilapia Using MobilenetV2 Convolutional Neural Network DOI

Israel F. Breta,

Karl Adriane D.C. Catalan,

Sev Kristian M. Constantino

и другие.

Опубликована: Июнь 29, 2024

Язык: Английский

Процитировано

0
FinSecure: Utilizing IoT Sensors for Formaldehyde Detection and Fish Freshness Detection for Enhancing Safety in Fish Consumption Using Machine Learning and Deep Learning DOI

S A Harish,

Komarellu Somesh,

Suntharavadivelan

и другие.

Опубликована: Июль 4, 2024

Язык: Английский

Процитировано

0
Nondestructive freshness prediction of large yellow croaker (Pseudosciaena crocea) using computer vision and machine learning techniques based on pupil color DOI
Xiao-Jing Wu, Qingxiang Zhang, Zhiqiang Wang

и другие.

Journal of Food Science, Год журнала: 2024, Номер unknown

Опубликована: Окт. 30, 2024

Abstract Conventional methods for evaluating of fish freshness based on physiological and biochemical are often destructive, complicated, costly. This study aimed to predict the large yellow croaker which was sampled every second day in 9 consecutive days at 4°C, using computer vision technology combined with pupil color parameters different machine learning algorithms (back propagation neural network, BPNN; radial basis function network; support vector regression; random forest regression, RFR). In process model building, RFR provided most accurate prediction value total volatile basic nitrogen (TVB‐N), R‐square test set () 0.993. The BPNN exhibited best fit predicting thiobarbituric acid (TBA), 0.959. Additionally, effective forecasting viable count (TVC), 0.935. After validation, root mean square error values TVB‐N value, TBA TVC were lowest, 0.764, 0.067, 0.219, respectively. It demonstrated applicability good predictive performance microbiological indicators. These findings also that monitoring changes could successfully chilled fish. Practical Application Scenario: Quality inspectors detect real time from beginning distribution selling site.

Язык: Английский

Процитировано

0
Fish‐Finder: A robust small target detection method for aquaculture fish in low‐quality underwater images DOI
Liang Liu, Junfeng Wu,

Haiyan Zhao

и другие.

Journal of Fish Biology, Год журнала: 2024, Номер unknown

Опубликована: Ноя. 25, 2024

Abstract Underwater fish object detection serves as a pivotal research direction in marine biology, aquaculture management, and computer vision, yet it poses substantial challenges due to the complexity of underwater environments, occultations, small‐sized frequently moving aquaculture. Addressing these challenges, we propose novel algorithm named Fish‐Finder. First, engendered structure titled “C2fBF,” utilizing dual‐path routing attention protocol BiFormer. The primary objective this is alleviate perturbations induced by intricacies during phase downsampling backbone network, thereby discerning conserving finer contextual features. Subsequently, co‐opted RepGFPN method within our neck network—a distinctive approach that adeptly merges high‐level semantic constructs with low‐level spatial specifics, thus fortifying its multi‐scale prowess. Then, an endeavor diminish sensitivity toward positional aberrations diminutive aquatic creatures, incorporated bounding box regression loss function, Wasserstein loss, existing CIoU. This innovative function gauges congruity between predicted Gaussian distribution reference distribution. Finally, regard dataset, independently assembled specific dataset termed “SmallFish.” unique meticulously designed for small‐scale intricate settings, includes 5000 annotated images small fish. Experimental results demonstrate that, compared state‐of‐the‐art methods, proposed improves accuracy , mean average precision (mAP) increases public Kaggle‐Fish SmallFish respectively.

Язык: Английский

Процитировано

0