High Speed Small Item Production Line Tracking Using Computer Vision and Cloud Computing

Counting high volumes of product and packages is prone to human error and is time consuming especially on a fast-moving conveyor system. To address this issue, automated counting systems are typically operated, the performance of which is dependent on the reliability of instruments used such as the sensors and scanners. Technological advancements in machine vision are progressing exponentially, making it increasingly accessible, cheap, and dependable for solving and executing real-time image related problems. In this paper, a vision-based high speed counting system is proposed. YOLOv8 was utilized as the machine learning model and is executed via Roboflow. Soap and small parcel items were focused on as a dataset. The generated mean average precision (mAP), precision, and recall for combined items is 98.6%, 96.8%, 96.3% with an inference and prediction time of 1:3 and 1:1 ratio respectively.