IoT Learning Device Recognition for Education Lab Management Using YOLO

Manual verification of IoT components in educational laboratories typically takes between 30 to 60 min per class. This paper introduces a YOLO11-based method for automating component recognition. We created a dataset consisting of 3,699 images with 4,734 instances across 30 different classes. Three variants of YOLO11 achieved a mean Average Precision (mAP) score of 0.767 to 0.776 on the test data. However, upon real-world deployment, we encountered significant domain shift, which caused performance to drop to a range of 0.008 to 0.018 mAP. By fine-tuning the model with domain-specific data, we were able to restore performance to between 0.728 and 0.748 mAP, representing an improvement of 40 to 88 times. This illustrates both the potential and the challenges associated with deploying computer vision technology in educational settings.