Real-Time Fish Detection and Tracking Using Yolov11 and Norfair

Object detection refers to the task of identifying andlocalizing instances of visual objects within digital images or videoframes, leveraging computer vision and deep learning techniques.In aquatic environments, this capability holds critical importancefor monitoring fish behavior, population analysis, and optimizingaquaculture systems. This project focuses on the implementationof a real-time fish detection and tracking system using YOLOv11(You Only Look Once), a state-of-the-art object detection modelknown for its high speed and accuracy, in combination withNorfair, a lightweight tracking library designed for fast, real- timeapplications. The proposed system is capable of processingcontinuous video streams, accurately detecting multiple fishspecies (with potential for multi-class support), and tracking theirtrajectories frame-by-frame using spatial consistency.A modular pipeline was developed, integrating live inference,object re-identification, and performance visualization through aninteractive Streamlit interface.This solution lays the groundwork for intelligent aquaculturemonitoring systems and can be extended to include behavioranalysis, abnormal pattern detection, and automated speciesclassification for large-scale deployments.