Video-Based Disease Detection in Vannamei Shrimp Using YOLOv8 Architecture

Abstract

Vannamei shrimp (Litopenaeus vannamei) is a high-value aquaculture commodity that significantly contributes to the fisheries sector. However, shrimp farming faces a high risk of disease outbreak to mass mortality and substantial economic losses. Conventional health detection methods rely on manual observation, which is subjective, inefficient, and requires expert knowledge. Therefore, this study proposes an automated shrimp health detection system based on video imagery using Convolutional Neural Networks (CNN) implemented through the YOLOv8 algorithm.The dataset consists of 2,000 images extracted from video frames of vannamei shrimp and categorized into healthy and diseased classes. The research methodology includes data preprocessing, augmentation, model training, and evaluation using performance metrics such as precision, recall, and mean Average Precision (mAP). The trained model is deployed in a web-based system using FastAPI and OpenCV to enable real-time detection. Experimental results show that the proposed CNN-based model achieves an mAP@0.5 of approximately 0.92 (92%), with precision and recall values of approximately 0.85 and 0.90, respectively. These results indicate strong detection performance under real-world conditions. The system is capable of automatically identifying shrimp health conditions and provides higher efficiency compared to manual inspection. This study demonstrates that deep learning-based computer vision has strong potential for early disease detection and can support sustainable aquaculture management