YOLO-TPS: A Multi-Module Synergistic High-Precision Fish-Disease Detection Model for Complex Aquaculture Environments

Area of Science:

• Aquaculture
• Animal Health
• Computer Vision

Background:

• Aquaculture is crucial for global food security but threatened by fish diseases.
• Manual disease diagnosis is inefficient and inaccurate, hindering sustainable aquaculture.
• Detecting small, varied lesions in complex environments is challenging.

Purpose of the Study:

• To develop a high-precision fish-disease detection model (YOLO-TPS) for aquaculture.
• To address limitations in small-lesion detection, size variation, and background complexity.
• To improve the accuracy and efficiency of fish disease diagnosis.

Main Methods:

• An improved YOLOv11n architecture was utilized.
• A multi-module synergy strategy and triple-attention mechanism were integrated.
• SPPF_TSFA and PC_Shuffleblock modules, along with a scale-aware dynamic IoU loss function (SDIoU), were incorporated.

Main Results:

• The YOLO-TPS model achieved superior performance: mAP0.5: 97.2%, Precision: 97.9%, Recall: 95.1%.
• Demonstrated enhanced multi-scale feature extraction and spatial awareness.
• Outperformed the baseline model in detecting fish diseases across six categories.

Conclusions:

• YOLO-TPS offers a robust and scalable solution for intelligent fish-disease diagnosis.
• The model significantly improves early-stage lesion detection and diagnostic accuracy.
• This technology supports sustainable aquaculture and effective animal health monitoring.