Artificial Intelligence of Things Real-Time Fish Counting System

This project aims to develop a real-time fish counting system using AIoT with three main objectives: developing a fish detection model using YOLOv8 and performing both quantitative and qualitative analysis, building a GUI for the Fish Counting System, and implementing real-time counting and analysis using IoT. The YOLOv8 model achieved 90.79% precision, 94.19% recall, 85.52% average precision, and a 92.46% F1-score, with real-world accuracy at 69.97%, an average processing time of 242.05ms per frame, and a rate of 4.13 frames per second. The GUI, built with Tkinter, includes Home, History, and Graph pages, allowing users to input video via IP Camera or local drive, with historical data stored in Google Firebase. Future work includes implementing DeepSORT for better fish tracking, using a polarizing filter on the camera, expanding the dataset to include more fish species and implementation using NVIDIA TensorRT.