This research demonstrates improved classification accuracy of fish feeding behaviors, suggesting efficient smart feeding systems via advanced computer vision techniques.
Key Points
Achieving a satiety classification accuracy of 98.1%, the method improves efficiency and model performance significantly.
The proposed two-stage network reduces parameter count by 31.4% and computational load by 26.2% while maintaining accuracy.
Utilizing satiety experiments, the study enables the generation of quantitative labels for fish feeding behaviors.
This novel approach addresses challenges in current methods, indicating a shift toward intelligent aquaculture strategies.