AI Based Product Object Detection & Counting, Sorting

Abstract

Object detection plays a crucial role in modern automation systems, particularly in areas such as retail inventor management, warehouse monitoring, and product categorization. Manual counting and classification of objects is labor- intensive, time consuming, and prone to human error. This project presents a YOLOv8-based real time object detection and classification system designed to detect, count, and categorize products into predefined classes such as Grocery, Medical, and Household items. The workflow includes dataset preparation, annotation, model training using YOLOv8, and post- processing steps for object counting and grouping. The YOLOv8 model is optimized for high- speed inference while retaining accuracy through fine tuning and anchor-free detection capabilities. Performance evaluation metrics such as MAP (Mean Average Precision), precision, recall, and inference time are used to assess model effectiveness. To enhance usability, the system provides real-time visual feedback with bounding boxes, labels, and product counts. Future improvements include integrating barcode recognition, implementing multi-camera tracking, and incorporating predictive analytics for demand forecasting.