Analyzing Fish Wellness Using Spatiotemporal Data & Behavior Recognition

Monitoring fish behavior is essential for fisheries research and aquaculture, as it plays a significant role in assessing fish well-being and applying effective husbandry practices. Although previous research has focused on fish tracking, there remains a need for post-analysis techniques that offer deeper insights into fish health and welfare. To address this, we present Fish-Eye, a system that combines tracking-by-detection using computer vision with image processing methods, emphasizing behavioral analysis. Our research highlights five key parameters: spacing, speed, direction, angle, and depth, providing a thorough understanding of fish behavior. In contrast to earlier studies that primarily utilized zebrafish, which have limitations such as smaller size, lower activity levels, and transparency, we select rainbow trout due to their more representative size and varied behaviors. Fish-Eye system involves meticulous data collection and the use of advanced software across multiple fish farms, contributing to the enhancement of fish welfare in aquaculture. Our comprehensive solution delivers valuable insights into fish behavior, improving monitoring and welfare practices. The complete implementation, including source code and additional materials, can be found in our GitHub repository. https://github.com/Vision-At-SEECS/Fish-Behavior-Analysis-for-Disease-Detection-using-AI