Program Type
Graduate
Faculty Advisor
Dr. Robin Ghosh
Document Type
Poster
Location
Face-to-face
Start Date
18-4-2024 9:00 AM
Abstract
Aquaculture expansion necessitates innovative disease detection methods for sustainable production. This study investigates the efficacy of Convolutional Neural Networks (CNNs) in classifying diseases affecting South Asian freshwater fish species. The dataset comprises 1747 images representing 7 class, healthy specimens and various diseases: bacterial, fungal, parasitic, and viral. The CNN architecture includes convolutional layers for feature extraction, max-pooling layers for down sampling, dense layers for classification, and dropout layers for regularization. Training employs categorical cross-entropy loss and the Adam optimizer over 30 epochs, monitoring both training and validation performance. Results indicate promising accuracy levels, with the model achieving 92.14% and test loss 0.2918.Training history analysis reveals an initial accuracy increase, followed by a plateau and eventual overfitting. Future improvements may involve regularization techniques implementation or additional data acquisition through data augmentation for better generalization performance. CNNs offer automated disease detection, alleviating the burden on aquatic animal health experts and facilitating timely interventions to mitigate infection spread and economic losses. Integration of deep learning enables real-time surveillance and adaptive management strategies, enhancing disease monitoring capabilities. This research aligns with broader efforts to improve global food security and underscores the transformative potential of AI-driven solutions in aquaculture disease management.
Recommended Citation
Tamut, Hayin; Rahaman, Musfikur; and Ghosh, Dr. Robin, "Enhancing Disease Detection in South Asian Freshwater Fish Aquaculture Through Convolutional Neural Networks" (2024). ATU Research Symposium. 27.
https://orc.library.atu.edu/atu_rs/2024/2024/27
Included in
Enhancing Disease Detection in South Asian Freshwater Fish Aquaculture Through Convolutional Neural Networks
Face-to-face
Aquaculture expansion necessitates innovative disease detection methods for sustainable production. This study investigates the efficacy of Convolutional Neural Networks (CNNs) in classifying diseases affecting South Asian freshwater fish species. The dataset comprises 1747 images representing 7 class, healthy specimens and various diseases: bacterial, fungal, parasitic, and viral. The CNN architecture includes convolutional layers for feature extraction, max-pooling layers for down sampling, dense layers for classification, and dropout layers for regularization. Training employs categorical cross-entropy loss and the Adam optimizer over 30 epochs, monitoring both training and validation performance. Results indicate promising accuracy levels, with the model achieving 92.14% and test loss 0.2918.Training history analysis reveals an initial accuracy increase, followed by a plateau and eventual overfitting. Future improvements may involve regularization techniques implementation or additional data acquisition through data augmentation for better generalization performance. CNNs offer automated disease detection, alleviating the burden on aquatic animal health experts and facilitating timely interventions to mitigate infection spread and economic losses. Integration of deep learning enables real-time surveillance and adaptive management strategies, enhancing disease monitoring capabilities. This research aligns with broader efforts to improve global food security and underscores the transformative potential of AI-driven solutions in aquaculture disease management.