Object Recognition using YOLOv8 for Car Seat Identification

This project aims to address the limitations of current techniques by enhancing the application of barcodes for car seat implementation in the automotive industry. The approach involves leveraging a comprehensive system of image recognition integrated with a cloud database. The system was deployed on a Raspberry Pi for efficient implementation. The primary objective is to reduce human errors, specifically in barcode labelling, by incorporating computer vision on the production line. Recognizing the error-prone nature of barcode labelling, leading to misplacements and challenges in identifying car seat types during storage, the project introduces object recognition as a secondary verification mechanism to minimize errors. Enhanced database capabilities enable efficient monitoring of the system’s output and facilitate an analysis of its effectiveness in car seat production. To optimize the system’s performance, we conducted experiments to assess the trade-offs, comparing the speed and accuracy of YOLOv8n, YOLOv8s, and YOLOv8m models. Subsequently, the YOLOv8n model with the average accuracy of 79.3% for type A, 83.3% for type B, and 81.1% for type C was deployed on a Raspberry Pi for practical implementation. Ultimately, the project aims to enhance the efficiency of the car seat recognition system, thereby reducing errors made by humans.