Enhanced recurrent capsule network with hyrbid optimization model for shrimp disease detection.

Area of Science:

• Aquaculture
• Biotechnology
• Computer Vision

Background:

• Disease detection is crucial for shrimp aquaculture sustainability.
• Early viral infection detection prevents significant economic losses.
• Current image processing models require improved accuracy for multi-disease detection.

Purpose of the Study:

• To develop a novel disease detection model for shrimp using an Enhanced Recurrent Capsule Network (ERCN).
• To enhance detection performance through a hybrid optimization model combining Harris Hawks Optimization (HHO) and Marine Predator Algorithm (MPA).
• To accurately classify various shrimp diseases using a single, improved model.

Main Methods:

• Utilized an Enhanced Recurrent Capsule Network (ERCN) with dynamic capsule routing and recurrent layers for spatial and temporal feature extraction.
• Incorporated spatial and channel attention models for optimal feature selection and fusion.
• Employed a dual-level feature fusion to combine local and global image features.
• Applied a hybrid optimization (HHO + MPA) to fine-tune classifier parameters.

Main Results:

• The proposed ERCN model demonstrated superior performance compared to conventional Recurrent Neural Network (RNN), Convolutional Neural Network (CNN), Gated Recurrent Unit (GRU), and Long Short Term Memory (LSTM) networks.
• Achieved high detection accuracy (95.2%), precision (94.9%), recall (93.5%), and F1-score (94.6%).
• Effectively classified different shrimp diseases using a single, integrated model.

Conclusions:

• The novel ERCN model with hybrid optimization significantly improves shrimp disease detection accuracy and reliability.
• This approach offers a promising solution for early and accurate diagnosis of multiple diseases in shrimp aquaculture.
• The findings contribute to enhancing disease management strategies and ensuring the health and sustainability of shrimp farming operations.